The Psychology of Survey Response

This valuable book examines the complex psychological processes involved in answering different types of survey questions. It proposes a theory about how respondents answer questions in surveys, reviews the relevant psycho­ logical and survey literatures, and traces out the implications of the theories and findings for survey practice. Individual chapters cover the comprehen­ sion of questions, recall of autobiographical memories, event dating, ques­ tions about behavioral frequency, retrieval and judgment for attitude questions, the translation of judgments into responses, special processes rel­ evant to questions about sensitive topics, and modes of data collection. Th� Psychology of Survey Response will appeal to ( 1) social psycholo­ gists, political scientists, and others who study public opinion or who use data from public opinion surveys; (2) cognitive psychologists and other re­ searchers who are interested in everyday memory and judgment processes; and (3) survey researchers, methodologists, and statisticians who are in­ volved in designing and carrying out surveys. Roger Tourangeau is Senior Methodologist at the Gallup Organization and has been a swvey researcher for more than 18 years. Lance J. Rips is Professor of Psychology at Northwestern University. He is the author of The Psychology of Proof and coeditor of Similarity and Sym­ bols in Hu1NJn Thinking_ Kenneth Rasinski is Research Scientist at the National Opinion Research Center, University of Chicago, where he conducts research in survey meth­ odology, substance abuse policy, and media and politics.

OGER TOURANGEAU The Gallup Organization

L

CE

J. RIPS

Northwestern University

KEN

ETH RASINSKI

National Opinion Research Center

.. :X:I,·.-• . � ,." : . ,..-·, · �II!

t.

'

�----� .

.

' '

.

.

. .•.

.

r

DGE UNIVERS....,ITY PRESS

I

PUBLISHED BY THE PRESS SYNDICATE OF THE UNIVERSITY . , T ·1he Pitt United ·

CAMBRIDGE UNIVER.SITY PRESS The au au. UK 40 Wat 20th Saeet, York, NY 10011-4211, USA 477 Wi Road Pon Melbourne, VIC 3207 Australia Ruiz de 13, 28014 Madrid, Dock House, Tbe Cape Towa 1001. South Africa ..

·

·

·

bbttp:l

org 1000

c

Ibis book is in and ro the proYisioas of , of any part I no of the··

to

•

take place without

•

,

.

statutory exupcion

·

•

•

·

First published 2000 Reprinted 2002, 2004 Printed in the United Stata of [BV)

l.ilwary of Congr•ss ��

·

.

Roa;er of aurvey

··

The Rips, p.

#

,

J.

u,

·

em.

lndudes ISBN

·

0-521-Sn46 0

(hb.). -ISBN 0-521-51629-(; (pbk.)

surveys polls - Evaluation. L Rips, m. Tide. 1.

2.

.

HN29.T68 300'.723

-

J. D.

Public opinion A.

2000

dell CIP

ISBN ISBN

0 521 57246 0 0 521 51629 6

OF

CAMBRIDGE

I

II

•

l

1 7

1.2 1.3 Theories

16

. pplica ion

II 2

of 2.

26 0

.3 2.4 2.

a

ts

..

on

2.6 2 6

•

.2

67

.3

ring

81 7

.4

'

..

..

Vll

Copynghted ma

nal

•

•

•

t

Vlll

ontent 4

00

4.

01

4.2

108

4.

21

4.4

urnrnary

36 5 5.3

Probability Judgnl nt

5.4

Conclusions

6

)63

ti ude Questions 6.1

'J'be Traditional View

6.2

I r a i Pa� t li f- rrpiing

6.

166 � "s � w �e� r

�

78

6.5 conclu ion. 7

t• ·tude Judgrn 7.1

2 l�7�

_

_ ________

1. 4 ts and Coot

Fortns of Context Effects 2 0

7.3

-

·

1

£ cti g the

nd

1

ir crion of

Context Effects 278 229 8

electin 230

Answers

2 9 4 eport1ng a •

9.1

.2 urtng •

rt1ng •

•

IX

10

9.

ill

9.

279

9.

28.6.

9..7

287

Mode of Data Collet:tion

289

1

ng

'·

Collection

290

10.2 '.[he Method of Contact and Adtninistration 10.3 Other Characteri tic ·.

·

..

f the

29!3

ata Collection

ethod

298

Data C.nllecrion Methods 10.5 Conclu ion·

ill 312 ill

0

15 Error 11.3 Impact on 1

.

ill ur ey

ractice·

4 Impact on P ychology

1 .5 Barri r , to Further ·

ccompli hment

323 335 337 M3

Atjthor ltzdex ubjea ltule.

ill

392

I

I

I • •

I Copy righted ma

ri I

To Karen, Julie, and Linda

This book examines surveys from a psychological perspective. It pro­ poses a theory about how respondents answer questions in surveys, reviews the relevant psychological and survey literatures, and traces out the implications of the theories and findings for survey practice. We hope the book appeals to a variety of audiences, including survey re­ searchers, methodologists, statisticians, and others who are involved in designing and carrying out surveys; political scientists, social psycholo­ gists, and others who study public opinion or who use data from public opinion surveys; cognitive psychologists and other researchers who are interested in everyday memory and judgment processes; and demogra­ phers, market researchers, sociologists, and anyone else who uses survey data and is curious about how such data come into being. Although we have written the book to be read from cover to cover, we recognize that not every reader will share our e·nthusiasm for all the topics the book includes. Readers who are most interested in public opinion may want to focus on Chapters 1, 2, 6, 7, and 8, skipping or skinuning the other chapters. Those who care mosdy about survey data on factual matters may want to focus instead on Chapters 1-5, 9, and 10. Those who are most interested in traditional issues in survey meth­

odology, such as question order effects and differences across methods of data collection, may want to concentrate on Chapters 4, 7, 9, 10, and 11. And those who are most curious about the cognitive psychology of

survey responses may want to focus on Chapters 1-5, 8, and 11. This book took shape over a period of four years. During that time, many people contributed to the book in many different ways, and we'd like to pause here to acknowledge their contributions and to offer our gratitude. •

XI

I

xii

Preface First of all, we thank the various undaunted souls who braved the early drafts of this book, hacking their way through thickets of tangled prose and negotiating mountains of conceptual confusion along the way. Their uggestions marked out our route through the later drafts of the book and have, we hope, blazed a trail that will be easier for later readers to follow. We start by singling out three good friends and careful readers - Stanley Presser, jon Krosnick, and Reid Hastie - who nude it through the entire rtunuscript and survived to give us their very useful comments. In addition, Fred Conrad and

orman Bradburn used drafts

of the chapters in courses they were teaching; they and their students gave us numerous helpful suggestions for improving the book. Other colleagues and friends read individual chapters or sections and also gave us useful feedback; these include Fred Conrad, Mick Couper, Bob Groves, Beth Proffitt, Michael Schober, Norbert Schwarz, and Michael Sht1m. We are grateful to all of these clear-sighted, kind, and tactful critics. Without their help, this book could have been a whole lot worse. There are also several people without whose help this book couldntt have been written at all. One of us (Roger Tourangeau) took shelter for three terms from his regular duties (first at the

ational Opinion Re­

search Center [NORC) and later at the Gallup Organization) at the joint Program in Survey Methodology (JPSM) on the catnpus of the University of Maryland. We thank Bob Groves and Stanley Presser for arranging this happy (.and much-needed) haven; our thank to

ancy Mathiowetz,

Martin David, and Mick Couper for their encoaragement during Tour­ angeau's stint at JPSM. In addition, we thank Phil DePoy and Kirk Wolter at

ORC and Max Larsen and Susan Nugent at Gallup for their

patience and support during Tourangeau�s various pan-ti111e leaves of absence from his day job. We are especially grateful to the Gallup Organization, which gave Tourangeau a partial subsidy during the final, critical birth pangs. Lance Rips also took a leave from his regular duties - as a profes or at

orthwestem University - to serve instead as a Fellow

of the Bureau of Labor Statistics (BLS) dwing the 1997-1998 academic year. It greatly speeded the completion of this book. We thank the BLS, the American Statistical Association, and the

ational Science Founda­

tion for sponsoring this fellowship. We especially thank Fred Conrad for his help during the fellowship period and Douglas Medin for suppo�rt from Northwestern. A project of this size and duration inevitably becomes something of a

trial to one's fatnily, and we'd like to thank ours for putting up with us during the past four years. We recogr1ize that forgetting curves, context

•

•

•

•

Xlll

effects, and response strategies are not exactly everyone's cup of tea. We are grateful to our wives for feigning, on many occasions quite convinc­ ingly, interest in these and other equally arcane topics that are not the stuff of fascinating dinner table conversation. We are grateful as well to our children, all of whom managed to stay out of jail and other 111ischief during this period of more than usual paternal preoccupation. Some of the heretofore unpublished data in Chapter 6 were collected under a grant from the NORC's Director's Fund. We gratefully ac­ knowledge NORC's support and thank Nor1nan Bradburn, NORC's director at the time, who arranged it.

CHAPTER ONE

troductio

and a Point of View

Survey research rests on the age-old practice of finding things out by asking people questions. In this respect, it has much in conu11on with a diverse set of activities ranging from police interrogations and courtroom proceedings to medical interviews and quiz shows. At the heart of each situation, one person asks another person questions for the p

e

of

obtaining information (Schuman & Presser, 1981). In surveys, this pair is the interviewer and the respondent. Interview­

ers can put questions to respondents face ... to-face, over the telephone, or , through a computer. However, the interviewer s questions and the re­ spondent's answers are always the central ingredients .. The question can ask about the personal activities or circ•J•nstances of the respondent (often called behavioral or factual questions) or they can seek the respon· dent,s opinion about an issue (attitude questions). The exam.ples in (1) are factual questions, and those in (2) are attitude questions from na­ tional s.urveys:

(1) a. Was anything stolen from you while you were away from home, for instance at work, in a theater or restaurant, or while traveling? { CS]1 b. Since the 1st of (month, 3 months ago), have you (or any members of your C[onsumer) U[nit)) received any bills for telephone servicesl Do not include bills used entirely for business purposes. [CE] 1 In citing examples from surveys, we follow the

phic convmtions of the s\U'Veys

themselves. We indicare the source of the survey in brackets after the example, usins the foUowing abbreviations for the five surveys from which we draw most of our examples: CE for Consumer Expenditure Survey, CPS for C urrent Population Survey, GSS for General Soc;ial Swvey, HJS for the U.S. Health Interview Survey, and CS for ational Crime Survey. 1

I

2

The Psychology of Survey Response (2) a. In general, do you favor or oppose the busing of (Negro/Black) and white school children from one school district to anotherl [GSS) b. Everything considered, would you say, in general, you approve or disapprove of wiretapping? (GSS]

The factuallattirude ter1ninology fits some survey questions more com­ fortably than others, but these examples give a rough idea of the rypes of questions that will concern us here. One might suppose that survey researchers would long ago have developed detailed models of the mental steps people go through in answering survey questions, models that spell out the implications these steps have for survey accuracy. After all, the accuracy of surveys depends almost completely on the accuracy of people's answers. Of course, with attitude questions, it is often difficult to decide what an accurate answer is. Still, answers to such questions documented response effects

-

prone to a variety of well·

are

differences in survey outcomes that reflect

seemingly irrelevant procedural details such as the order in which the answer categories are presented. These response effects may be due to problems in understanding the question, remembering relevant informa­ tion, producing an appropriate answer, or other mental processes. De­ spite their importance for understanding and evaluating surveys, the study of the components of the survey response process is in its infancy, having begun in earnest only in the 1980s. By conttast, rigorous mathematical formulations of sampling er1·ors have been available since the early 1950s, when the major texts on survey sampling appeared (Cochran, 1953; De1ning, 1950; Hansen, Hur­ witz, & Madow, 1953). Sampling error arises

use it is usually iln­

possible to interview everyone in the target population; thus, there is some uncertainty about the relation

een

estimates that come from

the interviewed respondents and the answers that might have come from the population as a whole. The study of survey sampling is devoted to reducing errors attributable to such effects, and it is both precise and scientific. But the study of questionnaire design, whose aim is to reduce response effects, remains an art (as Sheatsley, 1983, and Sudman & Bradburn, 1982, have noted). A major goal of this book is to provide a framework for understand­ ing response effects in surveys. We will propose a model of the survey response process and trace its implications for survey error. Our hope is to unify what is known about response effects by relating existing find-

3

IDtroductioa

ings to a cognitive model of the survey response process. Many reviews of the literan•re on response effects in surveys have noted the fragmen­ tary character of the research on this topic. We believe that a cognitive model of the survey response process can offer an improved map of this varied landscape. The tertllS response effeas, response process, and respondent mislead­ ingly suggest a behavioristic focus on the way that people respond phys­ ically to survey questions. We will continue to use these terms because they are universal in the survey literature, but we have already indicated that we do not share this point of view. Errors on surveys can be due to internal features of language comprehension, memory, and choice, as well as to the way people execute a response, and our theory builds on current research in cognitive psychology and artificial intelligence. We view response effects in surveys as a challenging

test

case for cognitive

science, one that goes beyond the simple tasks that typically find their way into the cognitive laboratory. In studying survey errors, we follow Newell's (1973) advice to consider large-scale domains that draw on 1nany mental abilities and that can therefore shed light on how cognition works as a unified system. We believe (and hope to persuade you) that a cognitive model of how people answer survey questions can offer as many insights to cognitive and social psychologists as to survey meth­ odologists.

1.1

Earlier Theories of the Response Process

Before presenting our own view of bow people answer survey questions, we consider two imponant antecedents: psychometric theories of atti­ tude measurement and early proposals within the survey methodology tradition. Comparing these theories to our own approach highlights what's distinctive to the point of view presented in this book.1 2

We concentrate in this

on historical antecedents, but contemporary ff!Rarch in

·

• about the way people answer survey questions, as will be evident in later chapten. For insta,nce, work · in AI has the process by which computers to answer be questioos framed in neryday lanpaac (c.a , Aile� 1995; Gracsscr, McMab� 8c john• • son, 1994; uhnert, 1978). The AI and .1· has concentrated on the difficult problem of identifying the spec:ific information that questions seek. So far, however, this · has had little di�ct impact on models of the survey .,. process. We attanpt to correct this oveni&ht in Chapter 2. ··

(AI) and

·

·

.

"

has also influenced

OW'·

•

·

..

·

·

I

4

The Psychology of Survey Response

1.1.1

Psychomettic Theories

Perhaps the first models of the survey response process were the ones developed by psychometricians - Gutttnan, Guilford, Likert, and their colleagues. This research produced the first systematic attitude measure­ ment techniques, techniques that quantified the strength of a person's conviction in an opinion (e.g., How strongly do you favor inaellSing aid to the homeless?). Of these early pioneers, Louis Thurstone was perhaps the leader in describing the underlying psychological processes that made attitude measurement possible (see, e.g., Thurstone, 1927).1 Thurstone's goal was to develop lttathemarical models that described the outcome of comparisons among several stimuli, including several statements about an issue (e.g., Aid to the hom8less should be increased; Society already does enough for the homeless; The homeless are a nuisance). Thurstone's formal models rested on a psychological theory of judg­ ment: the idea that people represent stimuli as points or

·

·

s on an

internal dimension (e.g., the dimension of strength of agreement with a position). Thus, his psychological models distinguished several compo­ nent processes, including the identification of the dimension of judgment (or, as Thurstone referred to it, the psychologiazl continuum), the judge,s reaction to the stimulus (the discriminal process), the

t of scale

·

values, and the comparison of pairs of stimuli (the calculation of their discriminal difference). In the case of an attitude question, Thurstonian theory predicts that people compare their own position with what they take to be the position implied by each statement in the scale. The result of the comparison deterrnines which statements they endorse. For a variety of reasons, these psychometric models have not played a prominent role in discussions of survey error. One

jor problem is

that, even though their emphasis is statistical, the psychometricians' 111athemarical models are not easily translated into the terms that have come to dominate discussions of survey error; the survey error models are extensions of those used to analyze random sampling error, and they focus on factual questions, for which it is possible (at least in principle) to measure the accuracy of the answer (e.g., Hansen et al., 1953; Chapter 11 and Groves, 1989, on the differences

see

the survey and

psychometric approaches to measurement error). In addition, from our

be found in Frank Ramsey's (1931) work on measure­ ment of utility, although this approach did not join the mainatream in psychology until the 1950s.

·'Perhaps another antecedent can

I

Introduction

current vantage point, the earlier psychometric models are hamstrung by the absence of a detailed description of the judgment process. Although Thurstone's model presents one view of how people choose among re­ sponse options, it is silent on questions of bow they identify the psycho­ logical continuu1n on the basis of the survey question, how they recruit relevant attitude information from memory, and many other crucial

es.

tssu •

1.1.2

Cannell's Process Theory

The model proposed by Cannell, Miller, and Oksenberg (1981;

see

Cannell, Marquis, & Laurent, 1977, for an earlier version) was perhaps the first model of the survey response process to reflect the new cognitive outlook within psychology. Their model distinguished two routes to an answer, one based on relatively careful processing of the question and the other based on superficial features of the interview situation, such as the interviewer's appearance (see Cannell et al., 1981). Careful answers are, according to this model, the product of the five sets of processes in

(3): (3) a. Comprehension of the question; b. Cognitive processing (i.e., assessments concerning the information sought, retrieval of relevant memories, and integration and response formulation); c. Evaluation of the accuracy of the response; d. Evaluation of the response based on goals other than accuracy; e. Accurate responding.

We illustrate the arrangement of these steps in Figure 1.1. Cannell, Miller, and Oksenberg (1981) view these five sets of processes as more or less sequential, although their model explicitly allowed for the possi­ bility that respondents might cycle back to an earlier stage if they judged their preliminary answer not to be accurate enough. Prior to the fifth stage, when respondents give an accurate (or at least adequate) answer, respondents could switch to the parallel track and alter their answer based on relatively su�

cial cues available in the interview situation -

cues such as the interviewer's appearance or the implied direction of the question. Responses based on such cues were likely to be biased by

acquiescence (the tendency to agree) or social desirability (the need to present oneself in a. favorable light). The model by Cannell and his colleagues has many attractive features,

5

0\

• •

a.

WI

,.._of

��

lhe��of

:

b.

.,,n• ,of r

1n aa::uracy

of

c.

:

R'1

·R gtves answer based on adequate

of in tenn1 of

other gollls

of and

of oniMtballl

'fr• chobl of telpanH beled on �cues A.. wv nr

(.......

,

b. The

and

R

by: a

blat

b.

...

c. c.

-· · --

R'a belefl. va�.-. and

d. OIIMar

1.1. Cannell, Miller, and Oksenberg's model of the survey response process. The boxes along t�he bottom of �the figure

represent the processes needed for a careful

answ

er; those along the top rep resa1t the processes leading to an inadequate

answer. Adapted from Cannell et al. (1981). Copyright C 1981. Adapted with permission of jossey-Bass.

-

bias

Introduction and it has spawned many related approaches

7

( see Section 1.3). The

notion that respondents might take different routes to arrive at an an­ swer is an appealing one, and the specific routes in the Cannell model­ one based on systematic processing of the question and the other based on more superficial processing - have a number of parallels in the psy­ chological literature. For example� discussions of attitude change have identified central and peripheral routes to persuasion (Chaiken,

1980;

Petty & Cacioppo,

1984). Hastie and Park's (1986) distinction between memory-based judgments and on-line (i.e., situation-based) judgments also bears similarities to the two tracks in Figure 1.1. Another attractive feature of the model is its explicit concern with the respondent's moti­ vation, including such motives as the desire to provide accurate infor· marion, to appear agreeable, and to avoid embarrassment. From our viewpoint, the model suffers from two major drawbacks. The first is that, because the model never assumed a central place in Cannell's work, it was never worked out it much detail. The most complete exposition of the model (in Cannell et al.,

1981) runs no more

than three pages. Most of the research inspired by the model has focused on improving respondent motivation in a general way rather than on testing predictions regarding the model's specific components. The sec­ ond drawback is related to the first. It is the model's rather sketchy treatment of the cognitive processes involved in responding to a ques­ tion, which the model's second stage lumps together. By contrast, the model distinguishes several stages in describing what happens after the respondent derives a preliminary answer. The respondent evaluates the initial answer in terms of its accuracy, then in tern1s of its compatibility with other goals, and finally may modify or discard it based on these earlier assessments. The model seems to asstame that respondents could answer questions accurately if only they wanted to and concentrates on whether they decide to answer accurately or not. We favor a different emphasis.

1.2

A Proposed Model of the Response Process

We have organized this book around a model that divides the survey response process into four major components - comprehension of the item, retrieval of relevant information,

use

of that information to make

required judgments, and selection and reporting of an answer (Touran­ geau,

1984, 1987; Tourangeau 8c Rasinski, 1988; see also Strack & Martin, 1987). Table 1.1 lists each of these components along with

0

I

8

The Psychology of Survey Response TABLE 1.1 Components of the Response Process Componmt Comprehension

Attend to questions and instructions Represent logical forrn of question Identify question focus (in tion sought) Link key terms to relevant concepts

Retrieval

Generate retrieval sttategy and cues Retrieve specific, generic memories Fill in missing details

judgment

Assess completeness and relevance of memories Draw inferences based on accessibility Integrate ntaterial retrieved Make estimate based on partial retrieval

Response

Map judgment onto response category Edit response

specific mental processes that they might include. In describing these processes, we don't mean to suggest that respondents

•

.

·

'Y perfortn

them all when they answer a survey question. Although some processes may be mandatory, others are clearly optional- a

set

of cognitive tools

that respondents can usc in constructing their answer. Exactly which set of processes they carry out will depend on how accurate they want their answer to be, on how quickly they need to produce it, and on many other factors. In this respect, the theory presented in Table 1.1 resembles approaches to decision making that emphasize the array of strategies that people bring to bear on a problem (Payne, Bettman, & johnson,

1993). These processes are also not exhaustive. We suggest some addi­ tions to the list in later chapters when we take up the components in more detail. Each of these components can give rise to response effects; respon­ dents may, for example, misinterpret the question, forget crucial infor­ mation, make erroneous inferences based on what they do retrieve, or map their answers onto an inappropriate response category. Both psy­ chological and survey research provide ample evidence of the errors each component produces, and to understand how the response process can go awry, we need to take a closer look at them. In reviewing these components, we also preview the material we will cover in the rest of this book.

9

Introduction

1.2.1

Comprehension

Comprehension encompasses such processes as attending to the ques­ tion and accompanying instructions, assigning a meaning to the surface fonn of the question, and inferring the question's point- that is, identi­ fying the inforanation sought (Clark, johnson,

1985; Graesser, McMahen,

&

1994; Lehnert� 1978). Current research on the prag111atics of

natural language emphasizes the reasoning people must perform in order to grasp the full implication of a sentence (Grice, Sperber & Wilson,

1989; Lewis, 1979;

1986); research on conversation emphasizes the way

conversational parbters cooperate to shape their mutual understanding (Clark & Schober,

1992).

As survey researchers have known for some time, many reporting

problems arise because respondents misunderstand the questions. Re­ spondents' attention tllaY wander during an interview, and they 111ay miss pan of the question; in a seH-administered questionnaire, they may not notice essential instructions or, having noticed, they may not bother to read them. The question may be double-barreled, inadvertently asking two or more questions at the same rime. The question tnay include terms that are unfamiliar to the respondent or terms that are understood in different ways by different respondents. The question may be too com­ plicated syntactically or it may contain detailed qualifications that are hard to understand. Familiar ter111S may nonetheless be vague, and even seemingly clear categories (such as

siblings) may include borderline cases

be rnisclassified (stepbrothers). With attitude questions, a key step in the comprehension process is identifying what issue the question that

is about. Chapter

2 of this book examines the comprehension of survey

questions in more detail; it presents a more detailed theory of this com­ ponent and reviews what is known about comprehension problems in surveys.

1.2.2

Retrieval for Factual

Qu estio ns

The retrieval component involves recalling relevant inforntation from long-term memory. This component encompasses such processes as adopting a retrieval strategy, generating specific retrieval cues to trigger recall, recollecting individual memories, and filling in

..

I memories

through inference. Several characteristics of the recalled material and of the initiating questions can affect the accuracy and the completeness of this component Uobe, Tourangeau, & Smith,

1993). These include the

I

10

The Psychology of Survey Response distinctiveness of the events, the degree of fit between the terms used in the question and the events' original enc

.

·

·

·

g, the ntiJltber and quality

of the cues that the question provides, the source of the memory (direct experience or secondhand knowledge), and the length of time since the events occurred. Chapter 3 examines these issues.

1.2.3

judgment for Factual Questions

Retrieval often does not yield an explicit answer to survey questions. The question may concern the total nu1nber of visits to the doctor in the last six months, the number of hours worked, or the total amount spent on retail purchases during the reference period. If so, respondents must sum the individual events they recalled during the retrieval phase in order �o find the total nt11nber. The judgment component comprises the processes that respondents use

to

combine or supplement what they have

rettieved. There are at least five major types of judgment processes that may come into play: (4) a. Judgments regarding the completeness or accuracy of retrieval; on the process of retrieval; b. Inferences .·

c. Inferences that fill in gaps in· what is recalled; d. Integration of the products of retrieval into a single overall judgment; e. Estimates that adjust for omissio�ns in retrieval.

The first three types of judgment depend on the relation between judgment and retrieval. Type (4a) detertnines whether further retrieval is warranted and whether specific memories fall within the scope of the question. Some judgments of this type can be seen as an extension of the comprehension component as respondents attempt to implement their understanding of the nature of the events covered by the question. The second type of judgment involves drawing conclusions from features of the retrieval process; for example, when retrieval is difficult or sketchy, respondents may conclude that the events in question happened infre­ quently or long ago or never took place at all (Brown, Rips, & Shevell, 1985; Gentner & Collins, 1981; Tversky & Kahnetnan, 1973). The judgments of the third type are attempts to reconstruct what happened by inferring missing details, often based on what typically happens dur­ ing an event of a given type. Respondents undertake the remaining types of judgment to transfortrl the retrieved information into an appropriate answer. The fourth process is necessary because people often retrieve fragmentary infortttation that •

Introduction

they must combine to produce a single response. This combination may involve simple numerical averaging (Anderson, 1981) or more complex types of estitnation. The final process involves situations in which what

is remembered forms part of a larger estimation strategy. For example, to answer a question about a lengthy ti1ne period, respondents may recaU the number of events in a recent, easily remembered portion of the period and then extrapolate to cover the entire period. The processes in (4) usually operate on retrieved information, but so

judgment supplants retrieval entirely. As Reder (1987) has

·

argued, people sometimes answer retrospective questions by considering the general plausibility of a response. Respondents who have never heard the term health maintenance organization (HMO) can probably infer from that fact alone that they do not belong to one. People can bypass retrieval of s

J

I

.

·

.

c inforn1ation in such cases.

ents about Dates and Durations

Survey questions often ask about events that occurred within some specific time frame; for example, the National Crime Survey asks ques· tion (5):

(5) During the last 6 months, did anyone steal things that belonged to you from inside ANY

ca

r or truck, such as packages or clothingl (NCS] •

Question (lb) is a similar item. Because people have difficulty remember­ ing exact dates (Friedman, 1993; Thompson, Skowronski, Larsen, & Betz, 1996; Wagenaar, 1986), they may report events that took place before the specified reference period (i.e., the time period that the ques­ tion asks about- the last six months in question (5)). This sort of error is known as forward telescoping, and this phenomenon has been the subject of

·

·

by both survey methodologists and cognitive psy­

chologists since Neter and Waksberg (1964) first doc111nented it. Reporting errors due to incorrect daring seem to arise through several distinct mechanisms. People may make incorrect inferences about timing based on the accessibility (or other properties) of the memory, incor­ rectly guess a date within an uncertain range, and round vague temporal inforn1ation to prototypical values (such

as

30 days). We take up these

issues in detail in Chapter 4. judgo1ents about Frequencies

Another popular type of survey question asks about the number of times the respondent has engaged in some activity or about the rate of

11

12

The PsychoiOSY of Survey Response events that happened in some reference period. Question (6) is typical of the sort of item we have in mind: (6)

During the past 12 months (that is since many times did

see or

(date) a year ago)1 about how

talk to a rnedical doctorl (Do not count

doctors seen while a patient in a hospital.) (Include the

visits you

already told me about.) [HIS)

Individual visits to the doctor may be difficult to remember, but it is often possible for people to reconstruct what must have taken place (e.g., Means, Nigam, Zarrow, Loftus, & Donaldson, 1989; A. F. Smith,

1991; see also Lessler, Tourangeau, & Salter, 1989, on the reporting of dental visits). For example, people 1nay recall only the regular pattern of events and use this rate to make an estimate for the full reference period (Burton & Blair, 1991). Chapter S considers the judgment processes involved in answering such questions.

1.2.4

Retrieval and J

·

.

for Attitude Questions

When we have to decide our attitude toward an issue, we need to consult our memory for relevant infortnation. But what exactly do re­

spondents retrieve when they answer a question about their attitudes? What does it mean for a respondent ro give an inaccurate report about an attitude? Most attitude research seems to take it for granted that having an attitude means having a preexisting judgment about an issue or a person and that people au

. tically invoke these judgments when

answering a relevant question (e.g., Fazio, 1989); responses to attitude questions do nor, as a result, seem to present much of a problem or to require an elaborate cognitive theory. We regard this view as overly simple. Along with Fischhoff (1991), we assume that there is a continuum corresponding to how well articu­ lated a respondent's attitude is. At the more articulated end, the respon­ dent has a prefortned opinion just waiting to be offered to the inter­ viewer; at the less articulated end, the respondent has no opinion whatever. Between these extremes, he or she may have a loosely related set of ideas to use in constructing an opinion or even a moderately well­ fortrted viewpoint to draw on. Respondents probably do not have perfectly well-articulated opinions on all attitude questions that survey researchers pose. First, evidence from large-scale attitude surveys indicates that on any given issue - no r11atter how familiar - a substantial portion of the population simply

Introduction does not have stable views. As Converse (1964, 1970) and others have shown, attitude responses over time sometimes show seemingly random shifts at the individual level even when no clear trends show up in the aggregate. Moreover, survey responses can shift dratnatically in response to minor changes in question wording or order. Second, even if respon­ dents do have more crystallized views about an issue, these views may not lend themselves to a clear-cut answer to the question at hand. The survey item may ask about an aspect of the issue that the respondent has not thought about. For instance, an item on the GSS asks whether abortions should be permitted in the case of rape; this item 1nay give even ardent pro-life advocates reason to stop and reflect before they answer. The judgment processes outlined in (4) are also relevant to attitude questions. For example, people may moderate or withhold their judg­ ment if they feel that the information they possess is not sufficient (Yzer­ byt, Schadron, Leyens, & Rocher, 1994); they may base their attitudes on what's most easily brought to mind (Ross & Sicoly, 1979); they may use

stereotypes and sche1nas to fill in inforrnation they can't recall (e.g.,

Hastie, 1981); and they may combine piecemeal evaluations into a single assessment (Anderson, 1981). In Chapters 6 and 7, we present a detailed model of the attitude response process; we return there to the issue of when people base their answers on existing evaluations and when they base them on more specific considerations.

1.2.S

Reporting and Response Selection

The model's final set of component processes involves selecting and reporting an answer. We include two groups of processes here- mapping the answer onto the appropriate scale or response option and ''editing'' the response for consistency, acceptability, or other criteria. Even when respondents have a clear answer to report, it may not be clear to them how to report it. The response options offered by the question are sometimes vague: In the case of attitude questions, where is the exact boundary •

·

een ''Strongly a·

'' and ''Agree''? In the do­

main of factual questions, how often do you have to eat in a restaurant in order for the frequency to qualify as ''seldom'' or ''often,'? Beyond the difficulties respondents may have with particular answer categories or response fortnats, they may differ in their approaches to selecting an answer. More than one response option may present a reasonable an­ swer for a given respondent. Some respondents 111ay work hard to choose

13

14

The Psychology of Survey Response the best possible answer; others may be content to pick the first accept­ able answer they consider. Respondents may also differ in their will­ ingness to give an answer at all or to opt out of a question by saying they do not know the answer. Survey researchers have exa1nined many questions regarding formats for answer categories, including how categories should be offered, whether each one should be labeled, how the categories should be ordered. and whether ''Don't know'' should be included among them. We discuss these issues in more detail in Chapter 8. Surveys often venture into areas that people do not ordinarily discuss with strangers. For example, a ntJmbcr of national surveys ask about the use of illicit drugs; other surveys ask about abortions, preferred methods of contraception, conslatnption of alcohol, medical conditions, or other topics that may cause embarrassment or resentment on the part of re­ spondents. In fact, some respondents of survey questions - Who lives here?

y regard one of the most basic -

as an unwarranted intrusion

(Tourangeau, Shapiro, Kearney, & Ernst, 1997). It is apparent that respondents do not always answer such questions truthfully. In Chapter

9, we consider the processes that govern the respondents' level of candor in answering sensitive questions- how respondents weigh the risks and benefits of responding truthfully (and of responding at all), what risks concern them, and what characteristics of the survey can raise or allevi­ ate these concerns. Some respondents may decide in a rational way whether to respond truthfully or to give evasive answers; others may decide in an unconscious and automatic way, in accord with rules that originally evolved for dealing with other situations.

1.2.6

Suan1oary

The theory of Table 1.1

sets

out 13 cognitive processes that people

may use to respond to a survey item. Despite the sheer number of processes, we should emphasize that survey responses

not the prod­

uct of lengthy deliberations. On the contrary, respondents take less than 5 seconds to answer typical attitude questions (Bassili & Flercher, 1991;

Tourangeau, Rasinski, & D'Andrade, 1991). In any model of this sort, a question inutlediately arises about the sequencing of the processes. Are the four rnain components distinct, nonoverlapping stages or are they simply classes of related processes ? There is little evidence to help

us

decide this issue, and theoretical argu­

ments can be made for either position. On the one hand, it

likely

Introduction

1S

that for many questions respondents methodically follow the logical sequence of first comprehending the item, then retrieving the relevant facts, then making whatever judgment is called for, and finally reporting

that judgment in the appropriate format. After all, how would respon­ dents know what to retrieve before they have understood the question? How could they n1akc a judgment if retrieval has not produced the necessary input? How could respondents report an answer before they have arrived at it? On the other hand, it is equally clear that there can be many varia­ tions in the response process. In the first place, there will often be at least partial overlap among components. For example, retrieval is likely to cotnmence before comprehension is complete; the very act of under­ standing key words in interpreting a question may trigger the spread of activation that is thought to be a central mechanism in retrieving mem­ ories (Anderson, 1983). In some cases, it may even make sense to think of retrieval as preceding comprehension; respondents may already have retrieved the inforrnation they need to answer one question in the pro­ cess of answering earlier questions. Similarly, judgment may parallel rather than follow retrieval. In answering an attitude question, respon­ dents may sequentially update their judgment about an issue as specific considerations come to mind. Or they tnay make judgments based on accessibility while they continue to retrieve memories. So any adequate model of the survey response process must allow for some overlapping processes. In the second place, the model must allow respondents to backtrack

from a ''later'' stage of the process to an ''earlier'' one. When retrieval yields little infor1nation, respondents may ask the interviewer for clarifi­ cation or try to reinterpret the question on their own. At least one class of judgments involves deciding whether additional retrieval is warranted. An adequate model of the survey response process must therefore allow for cycling between the judgment and rettieval components. Similarly,

selecting a response may require respondents to alter their j

so

that they can map the judgment onto one of the choices. Whenever the output from one component does not meet the requirement of another component, respondents may have to reexecute the earlier one. A third complication is that respondents can sometimes skip or trun­ cate a component. This accords with the cognitive toolbox approach that we mentioned in Section 1.2; respondents needn't employ all their response strategies in answering every question. Having understood a question as intrusive, respondents may skip retrieval and judgment alto-

I

•

16

The Psychology of Survey RespoDSC gether and go directly to response selection - offering an evasive re­ sponse or refusing to answer. In fact, inattentive respondents may even skip the comprehension stage and simply respond by saying they do not know the answer. With attitude items, respondents may omit the judg­ ment component once they have retrieved an overall opinion (or at least a definite impression), provided that the fonn of the question per1nits an answer based on an existing judgment. With factual items, retrieval may yield an answer that requires little

-

·

er cognitive processing; respon­

dents do not need to use much in the way of judgment to answer questions about their age, date of birth, or sex. And, as we have already noted, judgment may replace retrieval when people answer on the basis of plausibility or familiarity rather than specific memories. Because respondents

can

carry out components in parallel, because

they can backtrack from later components to earlier ones, and because they can completely skip components, it would be misleading to describe the four components as nonoverlapping stages. Although we suspect that comprehension- retrieval- judgment- reporting is the most corrunon order for the components, there is little evidence to support (or refute) this hypothesis, and other common patterns are worth noting.

1.3

Other Recent Proposals: High Road/Low Road Theories

The model in Table 1.1 constitutes an idealized list; the response process for a given item is likely to include only a subset of the processes identified there. Are there sequences other than the comprehension retrieval - judgment - reporting sequence that are likely to arise in practice? As it happens, several models of the survey response process identify such sequences as alternative paths to responding. We have already described the model proposed by Cannell and his colleagues in Section 1.1.2. This model features one track that includes most of the same processes that we have identified - comprehension, retrieval, judgment, and mapping - and a second track that modifies the response due to motives other than accuracy. This second track takes into account cue from the interview situation itself (see Figure 1.1). We have also discussed some reasons for preferring a different model from that of CanneD and his coUeagues. Here we consider an additional issue- the question of whether there are two distinct routes to arriving at an answer. It is certainly possible that, for some items, the conscien­ tious and superficial routes represent sharply distinct tracks consisting of clearly distinguishable processes; however, in many cases the distinction

between these two modes of processing is likely to blur. For example,

I

•

IDtroduction

17

even when respondents are trying to be conscientious, prior items can affect their answers; indeed, the effects of retrieving information relevant to one item may have an automatic effect - an impact outside the respondent's awareness or control - on retrieval for related questions later on. Similarly, respondents may use both information retrieved from memory and cues from the situation in formulating their answers. There is no reason, in principle, why both sources could not contribute infor­ mation to a response. Perhaps it is best to view Cannell's two tracks as two extremes on a continuum of processes that vary in the depth and the quality of thought that respondents give to their answers.

1.3 .. 1

The Satisficing Model

Two other models of the survey response process have appeared in the last few years, and they share with the Cannell model the assumption of dual paths to a survey response - a high road and a low road. The first is the satisficing model that Krosnick and Alwin have presented (Krosnick & Alwin, 1987; see also Krosnick, 1991). K.rosnick and Alwin distinguish between respondents who who

optimize

SIJtisfice

(the low road) and those

(the high road) in answering survey questions. In their

view, satisficing is not so much a strategy for choosing among response options as an overall approach to answering the questions (cf. Touran­ geau, 1984). Satisficing respondents do not seek to understand the ques­ tion completely, but just well enough to provide a plausible answer; they do not try to recall eve

·

g that is relevant, but just enough material

on which to base an answer; and so on. Satisficing thus resembles the more superficial branch of Cannell's two-track model. Similarly, opti­ rnizing respondents would seem to follow the more careful branch. Like the Cannell model, Krosnick's satisficing theory makes a sharp distinction among processes that probably vary continuously. Respon­ dents may process questions to differing depths, and they may not carry out each component process with the same level of care. There is no reason to assume that respondents who are inattentive while reading a question will necessarily do a poor job of retrieval. The key point is that respondents may carry out each cognitive operation carefully or sloppily.

1.3.2

Strack and Ma1·tin's Two-Track Theory

Strack and Martin (1987) have also proposed a two-track model, one that focuses on the response process for attitude questions. The two

I

18

The Psychology of Survey Response •

route

they identify correspond to the distinction between responses

based on an existing judgment and those based on a new judgment that respondents derive at the time they answer the question. The route for new judgments closely parallels the model presented .here; the key pro­ cesses comprising that route are interpreting the question, accessing rel­ evant information, ''computing'' the judgment, and formatting and ed· iring the response. The other route leaves out the judgment step and replaces rettieval of more specific information with retrieval of a prior judgment. Although we

inly agree that processes resembling both

tracks occur in surveys, we believe that both of Strack and Martin's track can be seen as special cases of the more general model we pre­ sented in Section 1.2. There are several reasons for the choice of the general model over Strack and Martin's more specific one. Fir t, we believe there is little evidence that r

pondents retrieve either an existing judgment or more

specific beliefs about an issue but never retrieve both. In fact, in Chapter 6, we review e idence that a mix of specific beliefs and existing judg­

ments may be the most common output of the retrieval process for attitude questions (in line with Fischhoff's, 1991, partial perspectives philosophy). ln any case, it seems unnecessary to exclude this possibility a priori. Second, we question the model's assumption that, when people retrieve an earlier opinion, they circumvent the judgment process en­ tirely. Although it may be possible to recall an attitude (e.g., that The

Last of the Mohicans is boring) without being able to remember any­ thing else about the topic, it is unclear how often this happens in surveys. ·

any attitude items require respondent to make new judgments even

when they retrieve a prior opinion. For example, agree/disagree items such as tho· e in (2) force respondents not only to retrieve their own views, but also to determine how close those views are to the position expre sed by the item. Respondents must still perform some sort of comparison proce s, however abbreviated. Like the routes in Figure 1.1, Strack and Martin's two branches represent pure cases. These ideal types are certainly worth noting, but it would be a mistake to see them as exhausting the possibilities. 4

.. Closer to our own approach is Schwarz's (1990) model for factual questions.. AJthough

this theory differentiates reports based on individual memories from those based on s

estimate

(a high road/low road difference)

it allows some variety in the

relations between judgm nt and retrieval processes .

I

Introduction

1.3.3

S111DID&ry

We prefer to think that quite a large number of paths to an answer are possible, depending on the effort that respondents are willing to invest and on the interplay between retrieval and judgment. In each case, the path traverses a subset of the processes identified here- which may be carried out well or sloppily, in parallel or in sequence, and with or without back

- as circumstances and motivation dictate.

A recent review by Jobe and Herranann (1996) describes seven models

of the response process, including sevetal we have already discussed. Some of the others encompass the same

listed in Table 1.1,

differing from the present model only in how they group these processes into larger components. For example, Forsyth, Lessler, and Hubbard

(1992) include separate comprehension and interpretation components; Willis, Royston, and Bercini (1991) include separate judgment and deci­ sion processc5. These differences seem more a matter of emphasis than of substance. Two of the models attempt to account not only for the mental operations of the respondents but for those of the interviewer as well (Esposito & Jobe, 1991; Sander, Conrad, Mullin, & Herrr11ann,

1992). 1.4

Applications of the Model

Models like the ones sut1unarized here are useful in part because they offer a new understanding of the sources of response effects in surveys and suggest methods for reducing such effects. For example, survey items such as (6) ask about the frequency of a speci6c class of events- visits to the doctor, days of illness, incidents of crime victimization, sexual parttters, and so on. The outcome of each component of the response process will affect the number of events reported in answer to such questions. Depending on how respondents understand the question- in particular, on how broadly or narrowly they define the class of events to be reported- more or fewer events will qualify. Similarly, the number of events they report will reflect the outcomes of retrieval and judgment: The more relevant events they can remember, the greater the number they will report, and the more events they see as falling within the reference period for the question, the more events will be included in the answers. Reporting errors are not the only place where cognitive models can shed light on survey problems. Chapters 7 and 10 of the book extend

19

20

The Psychology of Survey Response this model to two additional problems encountered in survey

- the

impact of the order of the questions and the effects of new computer­ assisted methods of data collection. One of the mo·st puzzling sets of findings in the survey literature concerns the impact of item context. In Chapter 7, we describe our theory of how earlier questions in the interview or questionnaire can affect each component of the response process (Strack & Tourangeau

·n, 1987·

Rasinski 1988). Prior items can change how respondents

interpret later question , what considerations they retrieve in forlttulat· ing their answers, which standards or nor1ns they apply in judging the is ue, and how they report their answers. In addition, different mecha­ nisms that produce context effects influence responses in different ways. Prior items sometimes influence later response in the direction of consis­ tency but sometimes have the opposite effect, producing apparent incon· i tencie . Complicating matters further, the context of an item can

change the overall direction of the answers or it can alter the correlations between an wer to different items. Over the Ia t 25 years, the face of surv y research has changed dra­ matically as computers have supplanted pencils and clipboards as the survey interviewer's most indispensable tools. Allowing the computer to collect the data has had a variety of effects, some of which have been subjected only to cursory investigation so far. The different methods of collecting survey data - self-administered questionnair s, interviews con­ duct d by telephon

or face·to-face, computer-as isted telephone and

personal interview , automated self-aditlinistered questionnaires - differ along many dimensions. We propose an analysis that singles out three k y characteri tic

of the method of dat

collection - the degree of

impersonality it conveys, the perception of legitimacy it fosters, and the lev I of cognitive burden it imposes on the respondent (and interviewer). Chapter 10 re iew the evidence regarding the effects of the method of data collection

1.5

n the an wers obtained.

ImpI· cations of t e Model

In discu sing the relationship between philosophy and psychology, Wil· liam James on . remarked that ''metaphysics ...spoils two good things when she injects herself into a natural science.�, There are those who might argue rh t the attempt to apply concepts. and method from the cognitive sciences to issues of survey methodology has had equally un­ happy re ults. Innovation ba ed on this attempt, sometimes dubbed the

I •

I

21 CASM movement (for cognitive aspects of survey methodology), have

been widely aocepted within the federal statistical cotrununity, but there is also widespread scepticism about the value of this approach. In our final chapter, we consider the results that CASM has achieved so far. Any effort to apply findings from one field to the problems of another raises several questions: Are researchers addressing the right problems? Are they applying valid theories? Have they developed the right appli­

cations of the theories for the problems at hand? We consider these questions in detail in Chapter 11, attempting to evaluate both our spe­ cific model and the more general movement to use cognitive science to tmprove survey pracnce. •

•

AJ.though the emphasis in this book is on the applicatio,n of cognitive mod,els to problems in survey methods, the findings in this growing litcranJre clearly have something to offer the cognitive sciences in return. We single out several areas where survey findings have implications for cognitive theory. First, apart from survey-related studies, there have been few formal investigations of memory for everyday events, the happenings that are the stuff of daily life. Surveys routinely ask about consumer purchases, visits to the doctor, searches for jobs, hours spent at work, illnesses, hospital stays, courses taken in school, and a host of other daily occtJrrences. To be sure, a few landmark investigations of everyday memory have appeared within cognitive psychology (see Chapter 3), but these are meager compared to the large nu1nber of studies conducted with a view to improving survey reports about such issues. The survey-inspired studies on everyday memory have yielded divi­ dends bearing on several theoretical issues in the cognitive sciences. For example, some of the clearest evidence for the existence of generic mem· ories - idealized memories of a class of recurring events - comes from attempts to understand survey reporting on dietary intake (A. F. Smith,

1991) and visits to the doctor (Means et al., 1989). Similarly, investigations of how respondents answer questions about the frequency of everyday events have yielded rich insight into estima· tion processes and their role in filling in information that memory alone cannot provide. Herrtnann (1992) describes some 15 strategies that re­ spondents use in answering survey frequency questions, many of them involving estimation. To cite one exarnple, Burton and Blair (1991) explore a strategy in which respondents recall events during a recent period and then make a rate-based projection for the entire period in question. 'Ibis strategy is far closer to the estimation procedures em­ ployed by statisticians or engineers than to the heuristics chat have taken

I

I

I

22

The Psychology of Survey esponse center stage in discussions of frequency judgments within the psycholog­ ical literature. Investigations of rounding in survey reporting (e.g., Hut­ tenlocher, Hedges, & Bradburn, 1990) have shed further light on how respondents compensate for vague or incomplete memories. Since the time of Bartlett's classic work (1932), psychologists have acknowledged that memory involves both retrieval and reconstruction; the survey-based studies on generic memory, estimation, and rounding have added consid­ erable detail to this picture. Another area where the survey literature has much to offer cognitive psychology involves what might be called proxy memory- memories for events experienced by other people. This topic has been almost com­ pletely neglected within the mainstream memory literature (see Larsert,

1988, for an exception) but has been a lively area within the movement to apply cognitive theories to issues of survey methods (e.g., Blair, Menon, & Bickart, 1991). In addition to its implications for the study of memory, the recent

efforts to apply cognitive theories to survey issues have much to contrib­ ute to the study of attitudes. As we've remarked, investigations of atti­ tudes in social psychology often seem to assume that respondents have a preexisting answer to most attitude questions and need only to read out this answer. The results from the survey literature present quite a differ­ ent picture: Responses to survey questions can become unreliable over rime (Converse, 1964, 1970) and show fluctuations as a consequence of seemingly 111inor changes in question wording (Schuman & Presser,

1981). In fact, simply changing the order of the questions can produce large swings in the answers (Tourangeau & Rasinski, 1988). If answers to attitude questions are simply readouts of stored judgments, it is not clear why question order should make such a difference- The study of order effects on responses to attitude questions has been a particularly fruitful area for the application of cognitive methods to a long-standing survey problem (Schwarz & Sudman, 1992; Tourangeau & Rasinski,

1988). These models are aimed at explaining survey results, but they have greatly expanded our understanding of assimilation and contrast effects in judgment more generally. In several areas, then, the effort to apply concepts and methods drawn

from psychology to problems in surveys has yielded benefits to both fields. Still, the sailing hasn't always been smooth. In Chapter 11, we consider some of the barriers to further progress.

•

CHAPTER TWO

•

·estlODS

Survey designers don't need to be reminded that the wording of the questions has an important impact on the results. Respondents can mis­ interpret even welJ..fortnulated questions, and when that happens, the question the respondent answen may not be the one the researcher intended to ask. Because of this obvious danger, the questions on na­ tional surveys are often subjected to empirical pretests. For example, the questionnaire designers may conduct cognitive interviews or focus groups in which they probe respondents' understanding of the questions and invite them to describe how they go about answering them (see Willis, DeMaio, & Harris-Kojetin, 1999, for a survey of these methods; we

·.

t a briefer discussion of them in Chapter 11). This practice is

useful in bringing to light problems the designers may have overlooked. This chapter looks at those aspects of survey questions that make them difficult for respondents to understand. These aspects are of many different sorts, ranging from features of gra1ru11ar and word meaning to the broader situation in which the respondent and interviewer find them­ selves. Grammar can come into play either because the sentence is struc­ turally ambiguous or because it includes complex clauses that respon­ dents cannot parse. As an example of structura l ambiguity, Item (1) asks respondents whether they agree or disagree with this statement: (1) Given the world situation, the government protects too many documents

by classifying them as SECRET and TOP SECRET. (GSS] As Fillmore (1999) points out, this sentence has two readings: According to one reading, the government, motivated by the world situation, pro­ tects too many doc\&Jnents; according to the other, the government pro­ tects more doc\JIDents than can be justified by the world situation. The ambiguity relates to syntactic structure (it depends on what part of the 23

I

•

24

The Psychology of Survey Respouse sentence the initial clause modifies), and it .

y affect how a respondent

answers the question. A recent study by Stinson (1997) centered on a question that illustrates the sort of gramatJa tical complexity that can lead to trouble: (2)

living where you do now and rneeting the expenses you consider nec­ essary, what would be the smallest incorne (before any deductions) you and your family would need to make ends meet EACH MO NTH ?

Although syntax can present hurdles for respondents, most studies of question wording have focused on semantic problems - problems of meaning- especially those involving the meaning of individual words. Many words in natural language are ambiguous (have more than one meaning) or are vague (have imprecise ranges of application). In addi­ tion, survey questions may include obscure or technical terms that are unfamiliar to respondents. Opinion surveys, for newly emerging issues that

are

···

pie,

·· ··

y ask about

unfatniliar to many of the respondents.

Vagueness and ambiguity can lead respondents to interpret questions in variable ways. For example, when Belson (1981) probed respondents in a follow-up interview about the meaning of Question (3), he found differences among respondents in the age they attributed to chil .·

­

dren: (3) Do you think that children suffer any ill effects from watching pro­ gramrnes with violence in them, other than ordinary Westemsl'

The respondents' difficulties might have been due to the ambiguity of the term children, which can refer either to sons and daughters of any age (as in How many children do you hllvef) or to youngsters in partic­ ular. Vagueness may be a more likely

source,

however,

use the

meaning of youngster is not crisply bounded (the division youngster and adult is not well defined). And, of course, the term ill effects is deliberately vague. SiJJtilarly, Belson found that respondents gave a wide range of interpretations to the adverbial quantifier usually in (4): (4) 1

For how many hours do you usually watch television on a weekday?

Questions (3) and {4) are not from aetna) surveys. Belson (1981) com them for research purposes in order to embody the types of question-wording problen1s most often ·

found in itenw provided by organizations whose representatives made available questionnair es which they had used over the past two yean" (p. 23). ..

I

•

Question (4) also illustrates a funher problem that affects comprehen­ sion. Many questions presuppose that certain characteristics apply to the respondent and then focus on an associated aspect. In (4), for example, usually presupposes that there is some usual pattern in the respondent's weekday television viewing, and the question focuses on how tnany hours per day make up that pattern. Presupposition and focus are nor­ mal components of a sentence's meaning, but they lead to difficulties in surveys when the presupposition fails to apply. If there is no regular pattern to the respondents' 1V watching, then they must either opt out of the question (e.g., by responding don't know) or reinterpret the ques­ tion in ways that apply to them. Difficulty with presuppositions may also occur in (5), another item from the General Social Survey, in which respondents have to rate their agreement or disagreement: (5)

Family life often suffers because men concentrate too much on their

work. [GSS] This item presupposes that men concentrate too much on their work and focuses on its effect on faznily life. A respondent who agrees with the subordinate clause (men concentrate too much) and disagrees with the 111ain clause (fa111ily life suffers) should have no special difficulty answer­ ing (5). But a respondent who disagrees with the subordinate clause may feel that the question doesn't properly apply to him or her (Fillmore,

1999). This item raises another difficulty: What sort of position would some­ one be advocating by making the statement in (5)? Is the intent to convey the feminist view that men should take on a fairer share of the household chores and child-rearing responsibilities? Or is the intent to convey a more funda1nentalist position that family life should take priority over outside activities? Depending on which reading the respondents give to the item, they may embrace or reject its implied sentiment. These examples illustrate the major classes of interpretive difficulty that survey designers encounter. The question's grammatical structure (its syntax) may be atr1biguous or too complicated for respondents to take in. Lengthy or complex questions can exceed respondents' capacity to process them, resulting in misinterpretations (e.g., just & Carpenter,

1992). The question's meaning (or semantics) may elude respondents if they misunderstand vague, unfamiliar, or ambiguous ternts or if they are misled by inapplicable presuppositions. Finally, the intended use of the question (its pragmatics) may create difficulties, as in (5). To

·n

exploring these comprehension difficulties more systematically, we begin

I

26

The Psychology of Survey Response by looking at the nature of questions and the processes involved in understanding them. The remaining section then exantine the contribu­ tions of granunar, meaning, and use in respondents' approach toques· nons. 41

2.1

at Is a

Question?

The comprehen ion difficulties that respondents face usually involve un­ derstanding que tions, and we will focus onquestions here. Obviously, respondent also hav to comprehend sentences of other

sorts,

especially

at the beginning of the survey interview, in explanatory passages, and during transitions between parts of the survey instrument (e.g., Now I'd like to ask you some questions about your children). With self­ administeredque tionnaires, comprehension of various kinds of instruc­ tion , especially those about the route respondents are supposed to take through theque tionnaire, can create problems as well U

& Dill­

man, 1997). But tnany of the aspects of comprehension that we will discuss in connection with questions carry over to other sentences as well. One imn1ediate difficulty in thinking aboutquestions, however, is that we can view them at different levels of analysis. Questions are associated with cenain surface forms generally give rise to a particular class of meanings, and are usually intended to perform a specific kind of action. But although these level

of forr11, meaning, and action are correlated

with each other, the correlation is far from perfect. We cannot concen­ trate on one level at the expense of the others. Like other complex linguistic objects,questions display a characteris­ tic granunatical and pho·nological structure. For exa1nple, questions of­ ten have inverted word order (Where was Catherine working last week? rather than C'Atberine was working where last week?) and a rising into­ nation contour. 2 Thus, the terttl question often refers to a class of lin2

But not always.. Echo questions

can

conversation (e.g., I fed your h . gerbil?). Subjea questions,

uch

as

preseJYe the order of a preceding statement in to th• gerbil. You fed my headband to the

Who fed your hudband to the gerbill,

normal word order (cf. I fed your headband to tbl! gerbil: 7.7). Rising intonation

likewise appears at the ends of

sec

some

also have

Radford, 1997, Section questions but not all.

Bolinger (1957, p. 1) cites the fotlo'W'ing example from Raymond Chandler, noting that the 6nal question probably doesn,t rise: Mr. Hady

is on nights and Mr. Flaclc on days. It's day now so it would be Mr. Flack

would

be on.

Where

can

I find him.

I

Survey Questions

27

guistic objects- interrogative sentences- that we typically use in asking for information. This meaning looms large in survey designers' talk of

question wording: Given the sort of infortnarion that we want, what's the best way of structuring the questio,n as a linguistic object to get at those facts or opinions? But once we begin to consider the possibility of alternative wordings, we seem to presume that there is something of which they are alternative versions: an abstract question that we can ask in different ways. If we're interested in finding out when someone

·

s his or her commute, we

might ask (6}, adapted from the Long Form used in the decennial census: (6} What time did Calvin usually leave home to go to work last weekl

But we could also use Could you please tell me when Calvin usually left

home to go to work last week? or On those days when this person worked last week, when did he

Of'

she usually leave home? These ver­

sions have clearly distinct linguistic forttts, but, at least in some situa­ tions, they get at the same information and should receive the same an­ swer. It is not easy to say exactly what the something is that each of these items expresses in common, but according to recent theories of the se.. mantics of questions (e.g., Groenendijk & Stokhof, 1997; Higgin­ botham, 1996), this shared aspect of meaning is what we will call a space

of uncertainty. This space consists of a set of possibilities, each of which constitutes a potential answer to the question. One of these possibilities is the correct answer. For Question (6) and its variants, this uncertainty space might be the set of all propositions of the fornt Last week, Calvin

usually left home to go to work at time t for all clock times t. A respon­ dent could give any of these propositions as an answer to the question, although only one of them (perhaps Calvin usually left home to go to

work at 9:15a.m.) would be the correct answer. We indicate this space in a schematic way in Figure 2.1a, where the different points in the space correspond to different answer possibilities and the starred point indi­ cates the correct possibility. If respondents also have response options for Question (6) (e.g., 7:00-7:59 a.m., 8:00-8:59 a.m., etc.), each option will correspond to a single possibility, collapsing the earlier set of points (see Figure 2.1b). If the response options are vague (e.g., morning, after­

noon, or evening), then the possibilities may share some of their proposi­ tions, as in Figure 2.1c. We discuss this last case in Section 2.4.2. An inteuogative sentence is a conunon way to express an uncenainty space, but it's not the only way. We can ask the same (abstract) question using an imperative sentence (Please tell me when Calvin usually left

home for work ltut week) or a declarative sentence (l,d like to know

I

28

The P ychology

What ti

did

of Surve y Response

Calvin usually leave home to go to work last week? •

Space a: ·-

- ----

--------�-- -

-

-

1 •

'

'

•

•

*CafWa

home at 9: 15 a.m. •

•

•

lll11 ;59 p.m.

C8.Mn

b: ,..---....-- ----=- ---....... ---

-...

--

Calvin 1e1tvee honMI at 7:00a.m.. •

•

•

•

Calvin

home at 7:59a.m.

catv In

home at8:00 am. • •

Catvin

home at 8:59 am.

C&tvin leaves home at 9:00 am. *

L_

• •

calvin leaves home at 9:59 a.m. -------

·--

- ------ ------

- ---=-

mom1ng

..

�-·-

•

*

·

.

.

..

-

--

• •

I

aftett•oon

•

-·

C&Jvin

at 11�30 a.m.

CalVin

home at 11:00 a.m.

1

I

• •

C81Yin

home at 5:30p.m.

Calvin leaves home at 5:00 p.m.

I

•

I

•

L_ -

--

.

·- -�--

·--

-----.....-----=-

-___J

2.1. A schematic view of the uncertainty space for the question

did Calvin usually leave home to go work last week?

When

Panel a shows the uncer­

tainty space without response options; panel b, for precise response options (6-

7 a.m., 7-8 a.m., etc)· panel c, for imprecise options

(morning, afternoon,

and

evening).

when Calvin usually left home for work last week). And interrogatives do not neces arily express an uncertainty space; they can express a statem,ent (Did you know that Calvin usually left home for work at 9:15 last week? How could you possibly think that I wiped my feet on

your mouse pad?) or request an action (Could you please stop wiping your feet on my mouse pad?), as well as laying out a space of possible answers.

Understanding Snrvey Questions

29

Finally, we can define questions not as a certain type of fortn or mearring, but as the activity that people perforln when they ask for information. Seen this way, questions are one sort of speech act (Searle,

1969). When someone asks when Calvin usually leaves home for work, he or she is usually making a request - that the listener provide infor­ mation about the time Calvin leaves home for work. If the meaning of a qu

·on is an

· ty space, then the request to the listener is o

provide information about which possibility in the space happens to be true. The standard way to make this request is to

use an

interrogative

sentence, and when that way is used we have the typical alignment of sentence form, meaning, and use that appears in Table 2.1 (adapting the view of Higginbothatn, 1996). In using the interrogative sentence (6), the questioner is expressing a space of uncertainty (that Calvin usually left home for work last week at time t for all relevant values of t) and requesting that the listener provide inforJttarion as to which of these possibilities is correct. As Table 2.1 also shows, this correlation

een

grartunatical fortn, meaning, and use for questions parallels a similar correlation for statements (including answers to questions). The correlation between form, meaning, and use in Table 2.1 is easy to break because interrogative sentences do not necessarily lay out

an

array of possibilities, as we've already noted. Likewise, one can use interrogative sentences without requesting information. Graesser, Huber, and Person (1991) distinguish four classes of gra.lniitatical questions, only one of which corresponds to a request for information of the sort found in surveys. The others monitor corrunon ground in conversation TABLE l.t Compon ents of Questions and Statements Use Questions

Answers

Interrogative sentence

Space of uncertainty

Requesting inforxnation

(e.g., "When did Cal­ ivn usually leave

(e.g., the set of prop­ ositions that Calvin

(e.g., requesting som one to infonn you

home to go to work

when Calvin usually leh

last week?")

usually left hometo go to work last week at timet)

home to go to work)

Declarative sentence

Proposition

Infortning (in response to a request)

(e.g., "Calvin usually

(e.g., the proposition

(e.g., asserting that Cal­

leaves home for work

that Calvin usually

vin usually leaves for

at 9:15,)

leaves homefor work at 9:15)

work at 9:15)

I

I

I

30

The P ychology of urvey Response

(Do you follow me?), coordinate social action (by iss1•ing instructions or seeking permi sion), or control conversation and attention (as with rhe­ torical question ). As Bolinger (1957, pp.2-3) put it, ''No one element a [question] .... For persons who demand rigorous

suffices to defin

definitions, the t rm question cannot be defined satisfactorily....

''

But despite this play in the connection between them, when inter�og­ ative ,

"nty paces, and requests for infort11arion line up, we have

something like a prototypical question. Although que tions

can

deviate

from the prototype in many ways, it provides a tarring point for our di

us ion of survey questions in the sections that follow.

Two Views of Comprehension: ltJllllediate Understanding versu Interpretation 2.2

Befor addre

tackling other a pects of question comprehension, we need to one oth

r

preliminary issue: What is comprehension? What is

the product of the question-understanding process? Unfortunately, the term comprehension is itself ambiguous. On the one hand, the meani g that we get from a word or a sentence must be relatively stable across people; how el e could we understand each other? But this stability implies that th interpretation of sentences has to be at least somewhat itrurtune to differences in the amount of knowledge about the concepts. When a survey includes a question about corttlttuting times, both the transportation planner who for'tnulated the qu who answer it must share some essential

cion and the cotrtntuter

t of meanings, even though

the two may differ in both the depth and kind of information they bring to bear on the concept commuting. But, on the other hand, it seems quite reasonable to think that a transportation planner attaches a much richer and more ab tract meaning to commuting than the typical com­ muter does. o how can their interpretations of a question about com­ muting really be the same? Thi discrepancy in our intuition about the tabil. ty of meaning across listener rnirrors a similar discrepancy in our intuitions about when we have successfully understood a sentence. In the normal course of a con· versation, we process sentences in a seemingly effortless way, and we feel we have int rpreted each sentence adequately as soon as (or perhaps a bit before) we come to the end. Unless we are brought up shon by a difficult grammatical c·onstruction (as in garden path sentences such as

The horse raced past the barn fell) or an unfamiliar word or phrase (say, computational lexicography), we understand the sentence immediately

u

Survey Questions

31

and are ready to move on to the next one. But it is also clear that comprehending a sentence doesn't always end at the period, with the reader or hearer secure in the right interpretation. If someone says that Bill gave another great sel'mon, and the hearer realizes that Bill is neither

a minister nor

a

priest, then he or she may interpret the sentence nonli­

terally - inferring that the speaker intends it ironically. But another hearer who doesn't know Bill might interpret the same statement liter­ ally. There is obviously plenty of room for misunderstaRding and contra­ dictory interpretations in ordinary talk.

2.2.1

Representation-of and Representatio.n-about the

Question How can we acconunodate these intuitions - that understanding is generally shared and inunediate but that it can reflect idiosyncratic knowledge and change or deepen over time? We assume that the product of comprehension consists of two parts, one obligatory and the other optional. Both patts are mental representations centered on the sentence that a person has just read or heard, but they differ in their content. One representation consists of a specification of the un,derlying grammatical and logical structure of the sentence, together with the lexical represen· tation of the individual words it contains. The other representation consists largely of inferences that the interpreter draws from the sentence in conjunction with other knowledge that he or he has available on that occasion. We call the first a representation of the sentence and the second a representation about the sentence (Rips, 1995}. The representation of the sentence is more or less constant across individuals competent in the language. The representation about the sentence varies however, de­ pending on the interpreter's standpoint, knowledge of the subject matter, knowledge of the speaker or writer, knowledge of the context in which the sentence was uttered or written, and probably many other factors. The representation-about will also vary with the amount of time and effort that the individual devotes to interpreting it: The greater the amount of interpreting that goes on, the richer this representation will be.

Consider, once again, Question (6), What time did Calvin usually leave home to go to work last week? According to some current theories

in syntax and semantics (e.g., Higginbotharn, 1996; Larson &

,

1995), the underlying structure of this sentence (its logical fortn) is similar to that in (6'):

I

I

I

•

32

The Psychology of Survey Response

• •

•

(6') [(Which(t)l I?[Last week� Calvin usually left home to go to work at t])],

where Which(t) specifies the questioned element of the sentence, ? marks the construction as a question, and t is a variable ranging over clock times. We discuss this type of formulation in more detail in the next

I

•

section, but for now we can assume that (6') gives the skeleton of the representation-of the question, the fra111ework that people compile as the result of hearing it. In addition, the representation-of the Calvin question must also contain some infor1nation about the meanings of the words and other lexical items in (6'). For exa1nple, the representation has to specify that Calvin is an expression that refers to an individual, that

work refers to an event, t to a rime, and so on.

•

I I

•

2.2.2

Constructing the Representation-of a Question

I

lf we step back from the formatting details, it is apparent that deriving

the representation-of a question involves several cognitive operations:

•

I

Representing the question in some format (like 6') that makes its logical structure clear;

•

•

picking out the question's focus ([Which (t)});

•

linking the nouns and pronouns to the relevant concepts in memory (e.g., associating the terms Calvin and last week with their cognitive representations);

•

assigning meanings to the predicates in the underlying representa­ tion (usually, leave home, go to work). •

Graesser and his colleagues include essentially these same operations in •

their model of the question interpretation process (e.g., Graesser, Bom­ mareddy, Swarner, & Golding, 1996; Graesser et al., 1994). What's controversial about the representation-of is its lexical content, the concepts that represent the meanings of the noun phrases, pronouns, and predicates. This component must suffice to determine the range of potential answers - the uncertainty space of the question - but beyond that point there is disagreement. According to some theories (e.g., Fodor,

1981, 1994; see also Anderson, 1983), the mental representations of the lexical items are fairly similar to words in natural language. According to others (e.g., jackendoff, 1991;

see

also Schank, 1975), the mental

representations are deeper and more fine-grained, speci

•

·

g both the

•

•

Understanding Survey Questions primitive conceptual elements that underlie words and the larger concep­ tual structures that these elements are embedded in.

2.2.3

The Representation-about the Sentence

Respondents do not stop interpreting a question when they have finished determining its representation-of. The question about Calvin's comn1uring time, for example, seems to imply that Calvin has some set pattern, a regular time when he leaves for work. For this reason, a respondent might infer that if Calvin works irregular hours, then the question doesn't apply to him. If the question is accompanied by a set of response options, the respondents may use the options to refine their interpretation of the question. If the response options are 6:00-7:00 a.m., 7:00-8:00 a.m., 8:00-9:00 a.m., 9:00-10:00 a.m., and ''other,'' then they know that the question doesn't require an answer that's precise to the minute. Likewise, they ntay take the response options

as

tacitly

specifying the usual range of answers that people give to such questions - the typical times people begin work. Perhaps they even assume that the actual frequency of starting times in the population is about equal for each of the response options. There is an endless set of possible inferences that respondents can make about the question that could be included ,in their representation· about it. Graesser and his colleagues distinguish 13 types of inferences that readers can make as they read a story (Graesser, Singer, & Tra­ basso, 1994). Only two of them (inferences that identify the referents of pronouns and noun phrases and those that assign case roles to the noun phases) are needed for the representation-of; the remainder all help elab­ orate the representation-about. Which inferences respondents actually make will depend on factors like the amount of ti111e they have to think about the question, their understanding of the purpose of the survey, the amount of inforrrtation they have about the topic, and

so

on. Although

some of these inferences might be more corrunon than others, it's un­ likely that every respondent will draw exactly the same ones. Thus, the representation-about the question is likely to vary across respondents and n1ay even vary for a single respondent across occasions.

2.2.4

Relation

een the Two Representations

Although we are treating the two representations as distinct entities, we do not mean to imply that there is no interplay

them.

33

34

The Psychology of Survey Response Certainly, people tll.af use the representation-of the sentence as the basis of inferences that become part of the representation-about it. And it is possible that the representation-about the sentence is involved in con­ structing its representation-of. As a person listens to a question, he or she may form hypotheses about how it will continue, hypotheses that may guide the construction of a representation-of the question. How­ ever, these hypotheses are not themselves part of the representation-of the sentence, and the listener may need to revise or discard them later when more of the sentence comes in. Thus, we needn't assu111e that people first construct the representation-of and then the representation­ about in strict sequence. Both the representation-of and the representation-about the question have an impact on respondents' answers, but these effects come from different directions. Complex wording or complex logical requirements can prevent respondents from being able to compute the representation­ of, and in such a case, respondents are in much the same situation they would be in if they had heard only a

t of the item. They are

missing basic information they need to deterntine the question's space of uncertainty and, as a result, they cannot be expected to come up with a relevant answer. Difficulties surrounding the representation-about the question, however, usually stem from too much information rather than from too little. Respondents may make unwarranted inferences about the question and

use

those inferences in constructing an inaccurate an­

swer. Suppose, for example, that respondents infer that the response options provide the typical answers to the question and base their own answers on whether they believe they are above or below average. Then their answers wiD vary with the particular set of categories the survey designer has chosen, no matter what the correct answer happens to be (Schwarz, 1996). In general, then, respondents' problems with represen­ tations-of a question may require clar· · · ·g and supplementing the ques­ tion itself. But problems with representations-about the question tnay require explicidy canceling inferences that the item the rema·

·

seems

to invite. In

sections of this chapter, we ntake use of this rep�esenta­

rion-about/representation-of distinction in examining effects due to the interrogative form, meaning, and use of survey questions.

2.3

Syntactic Difficulties in Question Wording

•

/

Let's return to the interrogative fortn, the first component of typical questions in Table 2.1, to see what difficulties it can pose for respon-

I

Survey Questions dents. In processing this component, the respondent's job is to get the question into its underlying propositional for111at (as in (6')) and to identify the question's focus. Difficulties in accomplishing these tasks partly reflect surface features of the interrogative forn1. In addition, they 111ay reflect syntactic ambiguity or excessive complexity.

Interrog ative For1o

2.3.1

As we noted earlier, interrogatives usually involve displacement of

words &om the positions they occupy in the corresponding declarative sentences. In yes/no questions (i.e., questions calling for a yes or no answer), these changes are srnall, involving a switch in the position of the sentence's subject and an auxiliary verb. For example, the interroga­ tive Have you had a mortgage on this property since the first of ]Mnei [CE) corresponds to the declarative You have had a mortgage on this property . . , where the auxiliary have has changed places with the subject you. When the declarative has no auxiliary, a forrJt of the word do appears instead at the nning of the question. Do you have a home equity loan? is the interrogative form of You have a home equity loan. Matters are more complicated, however, for questions that with wh-words, such as who, where, when, why, what, which, and bow. Most of these wh-quesrions shift the position of the subject and the auxiliary, just as yes/no questions do. For example, What would you have to spend uch month in order to provide the basic neassities for your family? flips the order of you and would. But the more dramatic difference is the position of the wh-word what. The corresponding de­ .

·

·

clarative seems to be of the form You would have to spend X ellch month . . . ; so the wh-word has switched to the front of the sentence from the position X occupies in the declarative version. In fact, there can be n�any embedded clauses separating the wh-word in complex questions from its corresponding position in a declarative. According to current generative theories of granm�.a r (see, e.g., Radford, 1997), wh-quesrions take shape through a process that moves the wh-word to the front of the sentence, leaving behind a silent (i.e., unpronounced) tical marker or trace in the original position. Thus, the representation of (7a) will contain a trace t• in the position shown in (7b): .·

.

(7) a. What would you have to spend each month in order to provide the basic necessities for your family?

35

36

The Psychology of Survey Response b. What would you have to spend r-each month in order to provide the basic necessities for your familyl c. I would have to spend

S 7,000 each month in order to provide the

basic necessities for my family.

The trick in understanding wh-questions is to deter1nine the trace posi­ tion, since it is the trace position that determines the question's focus­ rhe info

tion that needs to be filled in to answer the question. For

example, (7c) can serve as an answer to (7a), where $1,000 occupies the position of t• in (7b). In (7), the trace can occupy only one position, but questions can be ambiguous in this respect. The question in (8a), for instance, has two readings, depending on whether the question is asking about the time of the telling (trace ar the position shown in (8b)) or the time the telephone will be repaired (trace at the position shown in (8c)): (8) a. When did Lydia tell Emily the telephone would be fixedl b. When did lydia tell Emily t* the telephone

wou ld

be fixedl

c. When did Lydia tell Emily the telephone would be fixed r-l

How do people determine the position of the trace in understanding

wh-questions? Research on sentence parsing in artificial intelligence (AI) suggests some approaches to this problem (see Allen, 1995, Chapter 5, for a review). The basic idea is that when people encounter the wh-word at the beginning of a question, they store in memory information pre­ dicting that they will encounter a missing part in the remainder of the sentence. Which component will be missing is partly detern1ined by the nature of the wh-component itself. Because the trace shares the gram­ matical properties of the wh-phrase, the trace and the phrase will belong to related grammatical categories. In (7), for example, the missing piece will have to be a noun phrase because what fills the role of a noun phrase. The missing part of (8) will be a prepositional phrase, since when fills the role of a prepositional phrase (i.e., at what time)? As they process the sentence, people look for the missing component. If they

run

into a

stretch of the sentence that is ungranunatical because it lacks the pre­ dicted pan, they can plug the trace in to 611 the gap. H the sentence is grarrunarical without the predicted pan, they must nevenheless find a spot where the trace can go. Because monitoring for the trace position requires the listener to use working memory, this process is likely to make comprehension more difficult until he or she finds the correct location. (See just & Carpenter,

I

u

•

37

Survey

1992, for an account of the role of working memory in sentence compre· hension.)3 Psycholinguistic research suggests that people make guesses about the position of the trace in wh-atg\Jment questions as soon as they process the sentence's verb (e.g., Crain & Fodor, 1985). They

·

y even

make preli1runary decisions about trace positions before they reach any .

potential trace sites ( see Tannenhaus, Boland, Mauner, & Carlson, 1993). Consider how this trace-location process works in comprehending (7) and (8). In the case of (7a), the wh-word what alerts the listener that a noun phrase will be missing later in the question. As he or she processes the rest of the sentence, the listener encounters the word spend, a tran­ sitive verb that is missing its object (You spend is not grartmtatical). Since the expected noun phrase can fill the role of this missing compo­ nent, the trace must

occur

after spend, as in (7b). Parsing (Sa), however,

is a bit more difficult. One would expect a rnissing prepositional phrase

because of when, but there

are

no clues about the location of the phrase

in the rest of the question. Lydia told Emily the telephone would be

fixed is perfectly grammatical by itself, so there is no need for the prepositional phrase to patch up the syntax. Since the phrase could attach to either verb in the sentence (tell or would be fixed), the trace could appear in either the position it occupies in (8b) or the one in (8c). To decide between these two readings, the listener must rely on plausi­ bility, intonation, or other external factors (e.g., stressing tell- When did Lydia TELL Emily the telephone would be pxed?- favors (8b) over (8c)). As these exa1nples illustrate, wh-words differ in whether or not the component they represent is an obligatory or an optional part of the sentence. The words who(m}, which, and what often components called arguments that are required by verbs or prepositions in the rest of the sentence. The words where (in what place?), when (at what time?), why (for what reason?), and how (in what manner?) often begin components called adjuncts that are optional parts of the question. ArgtJtnent questions tend to provide more guidance than adjunct ques­ tions about where the trace should go. There is also evidence that ambig­ uous adjunct questions like (Sa} are somewhat easier to understand when ·

·

·

··

1

According to some parsing theories, grammatical rules determine the position of the trace automatically without · a separate monitoring process. We can assume, however, ·

· ·

that the rules that cany out this

draw

on

extra memory resources in

·

ing

these questions, so the effect on working n�emory may be the same.

I

38

The Psychology of Survey Response people interpret the trace as occurring in the main clause (i.e., (8b)) than when they interpret it as occurring in the subordinate clause (i.e., (8c)). In su1nmary, questions with different surface forms impose different demands on the listener. Relative to yes-no questions, wh-quesrions in­ crease the load on working memory, since they require the listener to reconstruct the position of the queried component. Among wh­ questions, those that concern arguments (questions about who, which, or what) may be somewhat easier to process than those that concern adjuncts (questions about where, when, why, or how). FinaUy, among adjunct questions, those that focus on an uncertain

of the rnain

clause are easier to understand than those that focus on an uncertain aspect of a subordinate clause.

2.3.2

Ambiguity and Complexity

The examples we have touched on so far have already indicated two funher sources of difficulty that can result from the syntax of a question - ambiguity and complexity. In the framework presented here, grammat­ ical ambiguity arises because the missing trace (the focus of the question) can be linked to more than one component of the underlying represen­ tation of the sentence. For example, in (8), the queried rime may involve when Lydia told Emily about the telephone or when the telephone will be fixed. In general, ambiguities arise in complex questions with multiple embedded clauses, so rewriting the question to eliminate the embedding seems a natural strategy for clarifying the intended meaning. Closed questions can also clarify things, since the answer options will point to the focal component of the question. Even when the complex form does not introduce syntactic ambigui­ ties, it may overload the processing resources of the respondents. Con­ sider this example from Fowler (1992), modeled on a question in the Health Interview Survey: (9)

During the past 12 months, since January 1, 1987, how many times have you seen or talked to a doctor or assistant about your health? Do not count any time you might have seen a doctor while you were a patient in a hospital, but count all other times you actually saw or talked to a medical doctor of any kind.

Both the syntax and semantics of the question are complicated. The question covers face-to-face and telephone consultations, with doctors

"of any kind"' as well as with ''assistants.'' In addition, respondents are to exclude such consultations if they took place whlle the respondents were hospitalized (or if they didn't concern the respondents' health) and to restrict their responses to visits that took place during the time frame specified. Processing this question into its underlying logical form is likely to impose quite a burden on working memory, one that may exceed the capacity of the respondent. Questions like (9) are the product of the practical constraints that questionnaire designers face. On the one hand, the questions need to specify the exact concepts the questions are trying to tap. In the case of

(9), most of the question and the accompanying instructions aim to define a particular set of medical visits - outpatient medical visits that occurred during a one-year period, where visit is construed broadly to include telephone consultations. On the other hand, there is the need to save time. The cost of a survey is, in part, a function of the length of the questionnaire. So, rather than ask a series of simpler questions to get at the same infor1nation as (9), the survey designers compress all four of the 111ain possibilities (face-to-face visits with a doctor, other face...to-face visits with medical personnel, telephone consultations with a doctor, telephone consultations with other medical personnel) into a single quesnon. •

In attitude surveys, there is another pressure that makes for compli­

cated questions. Many survey researchers believe that balanced items like (lOa) are better than items that state only one side of an issue (lOb): (1 0) a. Some people feel the federal government should take action to reduce the inflation rate even if it means that unemployment would go up a lot.Others feel the government should take action to reduce the rate of unemployment even if it means the inflation rate would go up a lot. Where would you place yourself on this (seven·point] scalel

b. Some people feel the federal government should take action to reduce the inflation rate even if it means that unemployment would go up a lot.What do you think? Do you agree strongly, agree, ...

Question (lOa) is taken from Converse and Presser (1986, p. 38), a widely used text on questionnaire design. The respondents are to indicate their answers on a seven-point scale, whose endpoints are labeled Reduce Inflation and Reduce Unemployment. A simpler alternative would be to present items like (lOb) and the parallel question on reducing unemploy-

40

The Psychology of Survey R ment; however, Converse and Presser observe that a substantial number of respondents would agree with both of these items, while (lOa) en­ courages such respondents to take a more definite position. Aside from the conceptual complexity of the underlying representa­ tion of the question, several other variables affect the load a question imposes on working memory (Just & Carpenter, 1992). One is the degree of embeddedness; questions along the lines of Is that the dog that

chased the cat that ate the rat that Karen saw? impose an especially heavy burden on processing capacity. Another is syntactic ambiguity, which increases the burden on working memory by forcing listeners to entertain two interpretations; garden path sentences (which require re­ interpretation at the end) are sirnilarly burdensome. A final variable is the individual respondent's working memory capacity. According to Just ·

and Carpenter (1992), individuals differ sharply in how much they can hold in working memory; questions that overburden one respondent may pose no particular problem for another. There are two main consequences to overloading working memory: Items may drop out of working memory (i.e., their level of activation 111ay get so low that the item can no longer be used in ongoing process­ ing) or cognitive processing may slow down Uust & Carpenter, 1992). Respondents may take a long rime to deal with Fowler's item on doctor visits in the past year, their representation of the question may omit some pan of the question's intended meaning, or both things may hap­ pen, with respondents taking a lot of time to come up with an incom­ plete interpretation of the question.

Setnantic Effects: Presupposition, Unfan1iliarity, and Vagueness 2.4

In the framework we have adopted, a question specifies an uncertainty space- a set of possibilities that correspond to the range of legitimate answers (see Figure 2.1). The survey designer's job is to ask a question in such a way as to convey the intended space, and the respondent,s job is to reconstruct the space and say where the correct answer lies within it. This perspective is a handy one because it allows us to discuss some of the common semantic problems that can derail a survey question. The question can express a space whose possibilities are not exhaustive, providing no location that the respondent can identify as a correct an­ swer. In the extreme, when the question uses terms that are unfamiliar to the respondent, the question may not express a space of possibilities

Survey Questions

u

at all. The question and its response alternatives can also produce a space that is poorly specified- for example, one in which the regions in the space aren't mutually exclusive. Lack of exhaustiveness results from (faulty) presupposition; lack of exclusiveness results from vagueness.

2.4.1

Presupposition

We found that statements like (5)- Family life often suffers because men

concentrate too tnuch on their work

-

can carry not-so-innocent

assumptions. The same is true of questions. Question (6) - What time

did Calvin usually leave home to go to work last week?- presupposes that Calvin had a job and asks what time he usually left to go there. If Calvin has no job, the question is ill-posed, and no direct answer is possible. Instead, the respondent would be forced to object that the question is simply not applicable to Calvin. Presuppositions like these arise because questioners must somehow describe the event or state about which they seek information. To answer the question, the ad­ dressee must identify the relevant events (e.g., Calvin's departures for work during the last week), identify the queried property of those events (their usual time of occurrence), and search memory for infor111ation relevant to the answer. (See Graesser, Robens, & Hackett-Renner, 1990, Graesser et al., 1994, and Singer, 1985, for theories of question­ answering that run along these lines.) Descriptive information in the question allows the addressee to perfor1n these tasks by specifying what the question is about. The presupposed information consists of things that the questioner and addressee normally assume to hold, since they a�e among the conditions that make the question meaningful. If the addressee does not agree with the information, however, then he or she must make some adjustment, either to accorrunodate or to reject it. We can think of a question's presuppositions as limiting the uncer .. tainty space that the question expresses. The uncertainty space that Question (6) expresses is roughly the set of propositions of the form

Calvin left home to go to work at t for all clock times t (e.g., 8:15 a.m.), as we saw in Figure 2.1. For the question to be correctly posed, some proposition in this space must be true. If none is true (because Calvin has no job) or if more than one is true (because he has several jobs or works irregular hours), then the question has no good answer. Presup­ positions restrict the range of possible propositions in the space: The more stringent the presuppositions, the narrower the range of allowable answers. At what time of the morning does Calvin usually leave home

41

42

The Psychology of Survey to go to work?

esponse

dd to the presuppositions of (6) that Calvin doesn't

work a late shift. The uncertainty space of the question in Figure 2.1 is correspondingly narrowed to just those titnes that

occur

dwing the

mom1ng. •

esearch by

oftus and her colleagues has documented the effects of

presupposition on people's memory for events. In these experimen , leading questions (i.e., qu stions containing a fal

e

presupposition about

an v nt) c n c use addressees to misremember the event as if tbe pre­ r

suppo ·irian

tru

. A question like How fast was the car going when

it went through the yield sign? can cause subjects to report the presence of a yield sign on a follow-up memory test, even if no such sign was part of the original traffic event that the subjects witnessed (Loftus, 1979). The cognitive basis of these false-memory effects remains controversial (for a taste of the controversy, see Belli, 1989, Tversky & Tuchin, 1989, and Zaragoza

McCloskey 1989), but it is clear that under certain

circumstances presuppositions can lead respondents to make incorrect inferences about

hat happened.

The effect of leading questions may be due in part to a norrnal reaction that occurs in conversation when a question contains a presup­ position about which the addressee has no prior knowledge. Presup· posed inform rion is something that the questioner assumes (and be­ lieves that the ad·dressee also ass u1nes). When the addressee does indeed know it, then all is well, and he or she can proceed to answer without further ado. n orne ca es, however, the addre see may not know the presupposed information (e.g., 1nay not know that Calvin has a job in the case of Question (6)). The presupposition is not necessarily rejected unless the addre

ee

believes it to be false. In fact, if the addressee thinks

the questioner is in a po ition to know the truth of the presupposition, the addressee may find it infortnative and remember it as factual. Al­ though this information is not something the addressee knew ·before, nevertheless the ddressee can accommodate to the presupposition, treat­ ing it as

tru

(

In the conte

Lewis, 1979, and Stalna

, 1974).

of a surv y, of course, respondents are unlikely to

suppose that the interviewer knows more about their personal circum­ stances than they themselves do, so they would be unlikely to interpfi t the interviewer'

presuppositions in a quesrio

like (6) as news. Some

survey questions, however, depend on more specialized information, such as fact about medical procedures and conditions (Have you or any family members ever suffered a myocardial infarction or heart attack? Do you belong to a health maintenance organization?) or knowledge of

I

I

I

I , I

•

u

Survey Questions

public issues (Do you favor or oppose the Agricultural Trade Act of

1978?). If the interviewer is asking their opinion about the Agricultural Trade Act of 1978, then respondents may infer that this is something they could (or should) have an opinion about, rather than an issue that is deliberately obscure or even nonexistent. How do respondents cope with questions that presuppose inforttta­ tion respondents do not have? With opinion questions, some respon­ dents (usually a majority) simply state that they don't know (Schutnan & Presser, 1981, Chapter 5). Those respondents who do answer may

look to the prior questions to support a guess about the meaning of the obscure issue (see chapter 7 for several examples of such context-based inferences with unfamiliar issues). When the item concerns a factual matter (Do you

or

any ntembers of you family have dental sealants?),

respondents seem to employ a variety of strategies (Lessler et al., 1989). They may ask for a definition of the unfamiliar ter111 or state that they do not know. The question's apparent presupposition that the respon­ dent ought to know the term may, however, encourage other strategies. (And, as we shall see, standard survey practice leaves interviewers little room to define unfamiliar ter1ns.) Some respondents seem to assimilate the problernatic term to a similar-sounding, more familiar one (denture

cream). Others conclude that the answer must be no, reasoning that they would probably be more familiar with the tern1 if it applied to them (Genmer and Collins, 1981, describe similar inferences based on lack of knowledge). Surely, we would know if we'd had a myocardial infarction. But each of these strategies for generating a substantive answer can lead to problems. Presuppositions are inescapable in natural language questions

use

they are a necessary part of expressing the range of uncertainty mat questions address. It is usually possible, however, to avoid most trouble­ some presuppositions through standard survey tactics. Questionnaires often include filter questions that route respondents around items that don't apply to them and issues they never heard of.. Similarly, it's com­ mon to add ''don't know'' or ''no opinion'' options in attitude questions to reduce the pressu1-e on respondents to fabricate opinions about issues they are not familiar with (Converse & Presser, 1986; Sudman & Brad­ burn, 1982). However, such tactics cannot eliminate all presuppositions. Even if we add a filter question that asks whether Calvin has a job, the main question about his commuting habits still presupposes that Calvin has only one job and that he leaves home for that job at a

·

time.

No question is presupposition-proof. The best any questionnaire can do

43

44

The Psycbolosy of Survey Response is to avoid presuppositions likely to be false in a significant number of cases within the intended population.

2.4.2

Vagueness in Questions and in Response Alternatives

Like presupposition, vagueness is impossible to avoid in natural lan­ guage. Vagueness occurs when it is unclear whether or not some descrip­ tor applies to an object or event. In the case of Question (3), repeated as (lla) later, it is unclear whether the descriptor children applies to teens (or to older offspring); there is no fact of the matter that could ·decide this issue. We might take a step toward .

king (lla) precise by stipulat­

ing the age boundary, as in (1 tb). Although this cutoff seems somewhat arbitrary, it at least manages to elitninate borderline cases that could make (lla) problematic for some respondents. However, what about ill

effects, programmes with violence, and ordinary Westerns? To deal with ill effects, for example, we could try listing ill effects, as in (tlc), but in addition to the fact that some ill effects

·

y themselves be vague, it

would be very difficult to spell out all the ill e

·

s that are relevant to

the question. Ul effects are inherendy open-ended, so making the phrase precise by fiat means missing some clearly relevant symptoms. Perhaps the best we can do is to acknowledge this with an and-so-on at the end of the list, as in (llc). Much the same is true of violence in this context. We can go some way toward clarifying the concept by giving examples, as in (ltd), but this is hardly precise enough to settle all questions about whether specific incidents constitute violence. (11) a. Do you think that children suffer any ill effects from watching pro­ grammes with violence in them, other than ordinary Westemsl b. Do you think that

people under the age of 14 suffer any ill effects

from watching programmes with violence in them, other than ordi­ nary Westerns? c. Do you think that

people under the age of 14 suffer any ill effects

from watching programmes with violence in them, other than ordi­ nary Westernsl By ill effects I mean increased aggression in school or at home, increased nightmares, inability to concentrate on routine chores, and

so

on.

d. Do you think that

people under the age of 14 suffer any ill effects

from watching programmes with violence in them, other than ordi­ nary Westemsl By ill effects I rnean increased aggression in school or at home, increased nightmares, inability to concentrate on routine

I

•

'

I

·

chores, and so on. By violence, I

Survey Questions

;

45

n graphic depictions of individ-

uals inflicting physical injuries on others or on themselves, tions of individuals wantonly damaging property or abusive behavior and language to others, and

so

··

ic-

ions,

on.

Vague Concepts Like the earlier example about doctor visits (9), (ltd) would clearly overwhelm the working memory capacity of most respondents, raising the issue of whether the effort to achieve precision is worthwhile. The danger in vagueness is supposed to be that some respondents will choose one way to make a vague question precise, whereas others will choose a different way, leading to uninterpretable variability in the responses. In commenting on his respondents' understanding of (lla), Belson (1981, p. 182) ranarks that ''It is . . . weD worth noting that there was a high degree of variability in the interpretation of terms like 'children', 'ill , eHects', 'violence - such that respondents who offer identical choice of answer may well have been considering rather different aspects of the matter concerned.'' He reports, in fact, that only 8% of respondents understood the question as intended. But the danger in making efforts to clarify vague concepts is that it produces lengthy, complicated ques­ tions that are also hard to interpret. It is certainly possible that respondents to questions like (lla) some­ times adopt interpretations that differ radically from those of others. For example, according to Belson, some respondents understood children as kids eight years old or younger, whereas others understood children as those 19-20 years old or younger. Later, we will also

see

evidence that

people interpret vague frequency expressions (e.g., pretty often) in differ­ ent ways. Nevertheless, some degree of vagueness

seems

built into the

meaning of ixnportant concepts such as children and violence. If these are indeed the concepts we're interested in, then we cannot avoid impre­ cision entirely. In fact, as we argue in Chapter 6, pan of what it means to have an attitude is to construe an attitude object in a characteristic way; part of the reason why some people favor welfare spending and some oppose it is that they

see

the issue in different terms. Such differ­

ences in how attitudinal concepts are interpreted are partly what we seek to measure by asking attitude questions. Moreover, some of the evidence of variability in understanding may

be due in part to differences in the way respondents specify

·

when they are asked to do so after the fact (e.g., in cognitive interviews or follow-up questionnaires), not to differences in their immediate com-

I

46

The Psychology of Survey Response

I

prehension. It is quite possible that in computing the representation-of the question, respondents do not make a vague expression precise but deal with the expression in its own tern1s. It is only when asked to explain what they understood by violence or ehildren that respondents reach for more specific explications - part of their representation-about the question.• At that point, variability is unavoidable because there are many ways to draw arbitrary boundaries.

I

Since the early 1970s, research on categorization has stressed the gradedness of everyday categories, such as furniture or flower (e.g., Rips, Shoben, & Smith, 1973; Rosch, 1973). Subjects rate some members of these categories as being more typical than others (e.g., chairs are judged to be more typical as

· ·.

·rure than 1V sets are), and these typicality

ratings predict many other aspects of their responses to these categories. For example, people take longer to decide that atypical members belong to the category than typical members (e.g., it takes longer to judge that a 1V set is furniture than that a chair is furniture), and they are more willing to generalize from typical than from atypical members. Later research has made it clear that these typicality effects are not always due to vagueness about what counts as members of a category. A category member can be atypical without necessarily being a border­ line case (Armstrong, Gleitman, & Gleianan, 1983). A dandelion may be an odd flower, for example, but it is clearly a flower nonetheless. In such cases, gradedness of typicality or exemplariness does not necessarily entail vagueness about category membership. For many categories, how­ ever, gradedness in typicality and gradedness in membership go hand in hand. A patch of color intermediate between orange and red is not only an atypical red but also a borderline red. 1V sets are not just atypical pieces of ft1rttiture but also borderline furniture, since there are no tech­ nical facts about either furniture or TVs that would reveal their true category status (Malt & johnson, 1992}. For categories like these, at­ tempts to draw the boundaries sharply may be useful for certain pur­ poses, but they also falsify to some extent the nature of the categories

4

The situation in interpreting vague

·

would then parallel other instances in

which people must introspcxt about their own mental

(Nisbett & Wilson,

l9n�. In asking respondmts what they understood by uioknu when they first heard Question (1 ta). we maybe asking them to try robe linguists and to analyze the meaning of the

term

from a theoretical perspective. Since respondents typically have no training

the raults may be highly variable. lllis is likely to be especially true when the follow-up question occurs on the day foUowing the interview, as it did in Belson·s (1981) study ( Ericsson & Simon, 1984) .. in perfornung such an analysis,

•

Unders

• .

S11rvey Quesrioas

themselves. In crafting questions, we may want to distinguish categories that are inherently vague (e.g., children, violence, ill effeas) from cate­ gories whose h1zziness is due to other factors (e.g., respondents' lack of knowledge). Exp

tory comments, such as those in (llc} and (ltd),

may be more valuable with unfamiliar concepts than with inherently vague ones. .

Vague Vagueness affects nearly all facets of language in surveys- not only the content of the questions, but also the wording of the response alter­ natives. Bradburn and Sudman (1979, Chapter 10) called attention to the fact that surveys often give their respondents a choice among ordered natural-language categories (e.g., never, not too often, pretty often, very

often) that may not have exact or constant numerical equivalents. Scales of this sort include adverbial quantifiers for frequency (e.g., never, not

too often), probability expressions (e.g., very unlikely, unlikely, likely), and quantifiers for amounts (e.g., none, few, some, many). Most of the expressions on these scales correspond, at best, to a range of numerical values. But both the range and the central tendency sometimes depend on the typical frequency of the quantified event (Pepper, 1981), the other alternatives on the scale (e.g.,

ewstead, 1988), and group and individ­

ual differences among respondents (e.g., Budescu & Wallsten, 1985; Schaeffer, 1991). Moxey and Sanford (1993) provide a comprehensive review of such factors. To take an extreme example, the statement that earthquakes occur very often in California implies a very different objec­ tive frequency (perhaps once a year) from the statement that someone sneezes very often (perhaps once a day). Bradburn and Miles (1979) show thar very often seems to refer to a somewhat higher frequency

as

applied to incidents of excitement than of boredom. Thus, it may be difficult to compare numerically a pretty often response to one question with the same response to a different question (or from a different respondent). The data in Figure 2.2 (from Schaeffer, 1991a, Table 2) illustrate the difficulty with vague response categories. The results come from a survey of 1,172 respondents who were asked, ''How often do you feel . . . particularly excited or interested in something? Very often, pretty often, not too often, or never?'' and ''How often do you feel bored?'' If the respondents gave an answer other than «'never," they were then asked, ''About how many times a week or a month did you mean?'' (Bradbur1a & Sudman, 1979). The figure plots the mean numerical frequency that

47

48

The Psychology of Survey Respm1se

$1 ,. ,�

1

Strategy

Month

Day of Month

Recall of exact date Use of landanark event Use of temporal period Guessing Other

tO

10

21

21

3

21

26

30

10

28

12

32

53

8

37

56

47

2 9

41 11

5 9

21 8

(1-100 Days Old)

49 20

Year Old)

(1992a) study, for example, participants reported retrieving dates for about 21% of personal events, where the events spanned an interval of several years. The subjects in this experiment were people who kept personal diaries, and these diaries provided the stimulus items. Since diary keepers may be especially well attuned to dates, these figures may overestintate the normal rate at which people retrieve dates for life events. A hint that this is so comes from Friedman's (1987) study of people's memory for the date of an earthquake they had experienced nine months before. Only lOo/o of participants reported directly retriev­

ing the month or the day of the month of the event. When people do rettieve dates, however, their accuracy is usually good. Participants were correct on 73.5% of trials in Burt's study when they said they had rettieved an explicit date.

Despite the fact that people don't often retrieve exact dates, they still do fairly well in locating the rime of personal events. In Burt's experi­ ment with diary keepers, participants assigned the correct date (year, month, and day of the month) to events on only 5.3o/o of trials; yet the correlation between the reported date and the correct date was extremely high (r

=

.92). If participants don't remember exact dates, then how do

114

The P ychology of Survey

espouse

they achieve this level of accuracy?-4 One obvious suggestion is that, although people don't remember the exact date of every event, they remember dates (or can deterr11ine them accurately) for cenain landmark events and use these landmarks to date less important events nearby. In Figure 4.2, the start of a new job played this landtnark, role: The event to be dated (purchasing the car radio) connects causally to the new job, since the radio

as bought to make commuting to the job more pleasan

•

Because it's easy to assign a date to the beginning of the job, it is possible to work backw rd in time to date the radio purchase. Landmark could include individual events of special importance (a wedding, the birth of a child, the death of a family member or friend) and calendar-bound events such as holidays and birthdays (see Shtlll,l 1998, for an e item

ended discussion). Table 4.2 shows the most frequent

that two groups of (mostly freshmen) students produced when

asked to list up to six landr11ark events that had occurred to them in the last year ( hum, 1997}. Landmark events were defined for the students simply as special times that '«>stand out from the more ordinary or mun­ dane events that happen to us.'' The first column of percentages comes from a group of tudent tested in September 1 996; the second comes from

a

simi)

c

group tested in January 1997. Important inciden

for

heshm,en tend to be those associated with the transition hool and college (Kurbat, Shevell, & Rips

hi.gh

998; Pillemer et al., 1986,

1988), and many of the listed events concern high school graduation acceptance by college

moving to college, first day at college, and

so

on.

The remaining non chool incidents include attending a prom, beginning or ending a romantic relationship, receiving new of the death of a friend or relative, and going on a vacation. The only calendar-based event the e students listed was their birthdays. It's possible, though, that the la.nd­ marks people use in dating events may differ from those they think of in responding to thi more general question. The frequency with which people report using landtrt.arks to date events varie

cross studies from lOo/o (Burt,

992a) to about 30%

(Friedman, 1987· Thompson et al., 1996), as Table 4.

indicate . Part

" Caution is necessary here, howevert since not all tudies of memory for

dates reveal this

White (1982) reports correlations in the range .26-.40 between correct date and estimated date for events in a year of his own life. The

level of ac uracy. For example,

diff�ren e may be due to the choice of events (White sometime selected commonplace events to record). to the different range of frequency of th

dates (one year v . up to nine years) or to the

en ..

I

I

I

Dates and Duratio ns TABLE 4.2 Percentage of Students I.isting Individual Events

as

Landmarks % of Students Who Mention

Event •

I I I

High school graduation Moving from home to college Acceptance at college Vacation Birthday High school prom Broke up with boy/girl friend

Death of friend or relative

Met boy/girl friend

Students Tested in September

Students Tested injmuary

ss 54 30 21 20 16 13 8 2

44

40 27 10 s 9 6 9 8

Note: Sample sizes were 255 in Septernber and 262 in January. Source: Data from Shum (1997}.

of this variability is due to differences in the retention interval across experiments, since the use of landmarks decreases with the age of the event to be dated (Bt•rt, 1992a; Thompson et al., 1996). Some of the variation across studies, however, could also be due to the way in which the investigators put the question about dating strategies. Participants choose their response from a fixed list of alternatives (e.g., Did you date the event by directly

remem bering

the date, by using a related landtnark

event, by using the general period of time during which the event oc­ cu"ed, etc.?), and the alternatives offered differ from one study to the next. A conclusion conunon to these studies, however, is that landmarks produce better estimates of the correct date than any other strategy exan1ined, except for direct retrieval of the date ((Baddeley et al., 1978; Burt, 1992a; Thompson et al., 1996).

Te1nporal Periods Chapter 3 noted that several theories assume that temporal periods or sequences organize memory for autobiographical events (Barsalou, 1988; Conway, 1996). According to these theories, long-tettn memory indexes events according to broad time periods or streams of connected events. It seems possible that these same periods or extended events could be helpful in deterrnining the time of an individual event that falls within them. For example, if you can determine through internal cues

115

116

Th P ychology of Survey Response

that an event occurred while you were on sabbatical in Burma (and if you know the time of the sabbatical), then you can obtain bounds on when the event happened. Temporal periods other than strictly autobio­ graphical one can also serve this bounding function: The 1993 model year for cars and the surnmer season in Figure 4.2 both aid in narrowing the time within which the target event (radio purchase) happened. As in the case of landmark events, some of the temporal periods can be calendar-bound (e.g., fall academic semester), whereas others are specific to individual (e.g., the period when I lived in California). Srudie of the methods people use to date events suggest that temporal periods are a coatunon source of infortnation (see Table 4.1). In one of the studies that Thompson et al. (1996) describe, participants reported using time periods to date 37% of recent events (1 to 100 days old) and 56% of more remote ones (more than 1 year old). Use of time periods, unlike landmarks thus increases with the age of the event to be dated, perhap u of the greater fragility of more specific information. Thomp on et al.' experiments also suggest that time periods produce fairly accurat srimates: Participants who report using temporal periods to date event are 1 s accurate than tho e who directly retrieve a date or a landmark, but they are more accurate than those reporting any of the other strategies recorded, including counting the nnntber of intervening events using clarity of memory to estimate rime, or using information from the event itself (Thomp on et al. 1996, Table 7.2). Using temporal periods to date events, how ver, 1 aves a characteristic mark on the data. If a person's only knowledge of the time of an event is that it fall within a spec·fic interval, then there's a tendency for the person to select a date near the center of the interval as the best guess. ng of the The result is that events that actually occurred near the interval tend to r eive too recent a date, and events that occurred near the end of the interval tend to receive too remote a date. This type of bias ppears clearly in experiments in which participants know that the stimulus event are drawn from a specified period, such as the last year (Huttenlocher Hedges & Prohaska, 1988; urbat et al., 1998; Thompon et al., 1996· White, 1982). However, a micro version of the same ffect also occurs if participants can localize the event to a horter interval within the larger period, such as an academic semester or quar­ ter. Events from the beginning of the quarter receive too recent a date, and events from the end of the quaner receive too remote a date (Hut­ tenlocher et al. 1988· Kurbat et al., 1998). There is a si1nilar phenome­ non in estimates of duration and of elapsed time, with longer intervals .

·

··

·

.·

·

I

I

I

Dates

Durations

underestitnated and shorter ones overestimated. All these related effects should probably be put down to people's general heuristics for selecting numbers from a bounded scale, and it is sometimes called a response

contraction bias in the psychometric literature ( see Chapter 8 and Poul­ ton, 1989). Of CO\Irse, landmark events and temporal periods don't exhaust the inforntation people can use to infer the rime of an event. For example, they can use elapsed time to estimate the time of occurrence because of the logical interdependence between the two that we noted earlier. This happened in Figure 4.2 in dating the start of the job three years ago. We will

ren1rn

to estimates of elapsed tin1e in Section 4.2.3, but the

set

of

strategies that people can use in reckoning dates is probably open-ended.

ory Information for Duration Despite the close logical relationship between time of occurrence and duration, research on these topics has proceeded independently. Labo­ ratory experiments on duration present participants with intervals of varying lengths (usually less than 10 minutes) and ask them to judge length in ordinary temporal units, to compare two lengths, or to repro­ duce the original length ( see Allan, 1979, and Zakay, 1990, for reviews). For example, a participant might hear auditory start and stop signals bounding a first target interval, a brief pause (the interstimulus interval), and then a second target interval. The participant must then decide whether the first or the second target interval was longer. The goal is to account for variables that affect the accuracy of participants' judgments. These variables include the lengths of the target intervals, the length of the intersrimulus interval, and the characteristics of events that occur within the intervals. Participants in these experiments either learn before­ hand that they will make temporal judgments (a prospective or inten­ tionol condition) or learn about the task after the stimulus presentation (a retrospective or incidental condition). For our purposes, retrospective judgments

are

the important ones, since survey respondents rarely know

prior to an event that they will later have to report its duration. We also concentrate here on studies of intervals that last for more than one minute, since the processing of very brief intervals is less relevant to surveys and rraay rely on a different set of psychological processes. Estimates of the length of a temporal interval usually increase linearly with true duration (e.g., Waterworth, 1985). In some experiments, the intercept of the linear function is greater than 0 and the slope is less than

117

118

The Psychology of Survey Response 1, so that participants are overestimating short intervals and t1nderesti· mating long ones. 'lois phenomenon tisne researchers call Vierordt;,s law ( see Woodrow, 1951). It is controversial, however, whether Vierordt's law reflects an panicipants' bias to respond with esti­ g more mates near the center of the stimulus range when they are uns•a1·e of the correct answer, another forrn of response contraction. een real duration Most of the relevant research on the relation and estimated duration has occurred under prospective conditions, since it is difficult to obtain enough rettos judgments to plot a psycho­ physical function. It would be useful to know whether retrospective judgments are also linear in real time. We also lack information about how estimated duration varies with length of the retention interval. Investigators seldom vary the retention interval in these laboratory ex­ periments, probably because they tend to view them as studies of tune perception rather than time memory ( see Schab & Crowder, 1989, for one exception). Loftus, Schooler, Boone, and Kline (1987) repon greater overestimates of the length of a videota bank robbery when partici­ .· ts 1 and 2) than pants gave judgments after a 48-hour delay ( inunediately after the ·. (ExperirJtent 3). However, Loftus and her colleagues did not vary the retention interval within a single experiment, and their three experiments differ in other potentially relev.ant ways besides their retention intervals. Investigators have adopted a nun1ber of theoretical positions to ex· plain the results of these duration experiments (e.g., Block, 1985; Om­ stein, 1969; Underwood, 1975). However, one thread that runs through . in these accounts is that the greater the DlJtnber of subjective an interval, the longer people judge it to be. The best-known finding of this sort is the �lied-duration illusion: The larger the number of auditory or visual signals that punctuate an interval, the longer it seetns (e.g., Ornstein, 1969, Experiment 1; see also Thomas & Brown, 1974, for brief intervals). Similarly, participants esti1nate the duration of a list of words as longer if highly salient words partition strings of less salient ones (Poynter, 1983; Zakay, Tsal, Moses, & Shahar, 1994) or if partic­ ipants have to alternate study tasks within the list (Block, 1985). Other factors, such as the nt•rnber of items that participants can recall from the intervals, the complexity of the items, the variability in item spacing, and the attentional demands of the task, all have less reliable effects on duration estimates. The effect of segmentation appears to depend, however, on the char­ acteristics of the event. In one experiment (Boltz, 1995, Experiment 2), ·

·

·

·

·

·

·

· ··

Dates and Durations participants judged the duration of a 1V show that included commercial breaks. When the conunercials interrupted natural patterns in the ongo­ ing action, the greater the number of cortunercials the longer and the less accurate participants' estimates of the show's length. However, when the commercials appeared at the seams of the action, the number of com­ mercials did not lead to greater overestintates. A possible explanation for this result is that corlnr1ercials added further subjective

.·

ts only

if they did not coincide with the segments imposed by the action.

This

.·

tation effect may be closely related to (perhaps a special

case of) people's tendency to estimate partitioned quantities as greater than equivalent unparririoned ones. For example, Tversky and Koehler (1994) doc1ament a similar unpacking effect in probability estitnation: People estimate the probability of an exhaustive set of subcategories {e.g., the probability that Calvin died of heart disease, cancer, or other natural causes) as greater than that of the entire category (Calvin died of natural causes). (See Section 5.3.2 for further discussion.) Similarly, Pel­ ham, Sumarta, and Myaskovsky (1994) describe a numerosity heuristic: a tendency to esrirnate overall quantity (e.g., the sum of a set of numbers) from the nt11nber of pieces it encompasses (the number of addends). Such a heuristic might be responsible for the segmentation effect in retrospec­ tive judgments if participants fail to attend to specifically temporal char­ acteristics of the original events and therefore have to base their esti­ mates on ancillary infor1nation, such as the number of

·

ts they

remember. Segments may thus be among the stock of infortnation that people can use as a basis for temporal inferences. In addition, there are many (meta-) beliefs about the passage of time that they could also use to adjust estintates based on the nuanber of

. For example,

people tend to think that intervals filled with unpleasant activities, few activities, monotonous activities, or easy activities seem longer than those filled with pleasant, numerous, variable, or difficult ones (Galinat & Borg, 1987). Some research on autobiographical memory has also examined

whether these laboratory effects generalize to everyday experiences. As an analog of the filled-duration illusion, Bun (1992b) asked participants to estimate the duration of two groups of events that he had culled from the participants' personal diaries. One group consisted of ''filled events,'' each a continuous stream of related activitiest such as a hospitalization or an official trip; the other group consisted of ''unfilled events,'' each including two related, but temporally separat� incidents (e.g., receiving an invitation and attending a party, applying for a position and hearing

119

120

The Psychology of Sw,vey Response the result). The filled-duration illusion would

seem

to predict greater

signed error (i.e., estimated duration - true duration) for filled than for unfilled events, but the results of the study showed no difference in signed error and somewhat smaller absolute error (i.e., I estimated dura­ tion - true duration I) for the filled events.5 The parallels are unclear, however, between the filled-unfilled distinction in this study and in the earlier lab experiments. What 6lls the laboratory intervals are sequences of sounds or lights, not causally connected incidents. Moreover, the laboratory filler may exert its effect by dividing the overall interval into parts, and there is no guarantee that the autobiographical filler played the same role. In clarifying the relationship, it would be helpful to know the n•Jmber of perceived subevents within these autobiographical items. As Bu•t points out, the relatively small effects of the retention interval

n1ay be due to the fact that participants could use the description of the event as the basis of an estimate for its duration, even if they were unable to recall the event itself. When asked the duration of a specific business trip, for example, a participant could use the usual range of business trips to derive a reasonable guess.

4.2.3

Meo1ory for Elapsed Ti111e

These laboratory studies on perceived duration don't distinguish judg­ ments of duration (how long an event lasted) from judgments of elapsed time (how long ago it took place). Participants give their time judgments just after the end of the event, so elapsed rime is minimal and approxi­ mately constant during the experisnent. For naturalistic events, though, there are two studies that focus on elapsed ti111e - one by Ferguson and Marrin (1983) and the other by Huttenlocher et al. (1990). Ferguson and Martin's study concerned public events within the past five years; participants judged how long ago each event had occurred (e.g., ''How long ago was Pope john Paul ll's visit to the United States?''). The study by Huttenlocher and her colleagues was embedded in a follow-up phone call that took place up to 60 days after a General Social Survey inter· view. During the call, the interviewer asked respondents, ''How many days ago did the interview take place?" Both studies obtained an elapsed-time version of the usual Vierordt pattern: Recent events received estimates that were too far in the past, whereas older events received estimates that were too recent. As we ' For a description of these and other error measures, see Chapter 8.

Dates and Durations

noted in Chapter 3, survey researchers refer to such dating errors as telescoping, presumably because, in survey settings, the errors typically involve events reponed as happening more recently than they actually happened; from the vantage point of the interview, the events are seen as closer in rime than they really are. But errors in the opposite direction (backward telescoping) are made as well. Backward telescoping is more prominent in Ferguson and Martin's data, forward telescoping in the data of Huttenlocher and her colleagues. This could be due either to the difference in the range of dates (two months versus several years) or to the distinction between public and personal events. The time of a recent personal event, such as an interview, may be easier to discern than that of a recent public event. Our knowledge of most public events is proba­ bly less vivid initially than our knowledge of personal experiences and may level off more quickly. If so, lack of information about recent public events could allow more room for backward telescoping.6 According to Ferguson and Martin, people compute elapsed time from the time of occurrence of the event, which they encode during the initial experience; according to Huttenlocher and her colleagues, people compute elapsed time directly. Both sets of investigators, however, ac­ knowledge at least implicitly that elapsed time estir11ates probably de­ pend on both sources of information. According to Ferguson and Mar­ tin, people use related events to determine an approximate time of occurrence when no direct calendar information is available for the target item. Huttenlocher and her colleagues lean toward event se­ quences (see Chapter 3) to explain their respondents' preference for elapsed time over calendar dates in answering the question ''When did the interview take place?'' The model they propose for elapsed time, however, makes no essential

use

of these sequences; estimates of elapsed

time derive from time-of-occurrence infortnation, subject to rounding and to adjustment for the ends of the stimulus range.

4.3

Indirect Effects of Time on Sul'vey Responses

Difficulties people have in determining the time or duration of events show up in survey data as measurement error, often as unexplained ' Huttenlocher et al. (1990) also show that respondents round estimates to times that are stand·ins for calendar units (seven days, thirty days) or are multiples of five. Hornik ( 1981 ) reports a similar tendency to round to multiples of five in duration estimates. The multiples·of-five sttategy does not occur, however, when partici pants must produce time­ of nee answen rather than elapsed times {Skowronski ec al., 1994). Sec Chapter 8 for further discussion of roundin&. ·

121

122

The Psychology of Survey Respoase bunchings of reponed incidents: Plotting the data over the titne line reveals that events are overreported at certain periods and underreported at others in ways that researchers can't explain. In this section, we consider two uch phenomena - the seam effect and telescoping - that survey researchers have clearly doclJinented, and temporal inferences, like those we've

we

examine whether

discussed earlier, can help account

for them.

The placement of an event in time can have dra1natic effects on survey reports, even when the respondents don't have to provide an explicit date or time. 7 One example

occurs

in longitudinal (or

panel) surveys,

such as the Survey of Income and Program Participation (SIPP) and the Panel Study of Income

amics (PSID). These surveys interview each

respondent at fixed intervals (or

waves) and ask the respondent for infor1nation about the intervening period. SIPP, for exa1nple, interviews

respondents three times a year; each interview covers their

·parion

·

in the labor force and income sources in each of the preceding four months. An individual respondent might be interviewed at the

·

·

·

of June to report on employment and income for the months of Febru­ ary, March, April, and May; the respondent would be interviewed again in October and would report on june, July, August, and September; and so

on, for a total of seven interviews Uabine, King, & Petroni,

1990).

Figure 4.3a shows part of the interview schedule for one group of re­ spondents. (SIPP includes several rotation groups whose schedules are phase-shifted by one month.) A problematic feature of such surveys is called the

seam

effect

-

the

tendency for month-to-month changes in the data to concenttate suspi­ ciously in adjacent months that were covered in different interviews. For example, a SIPP respondent interviewed in October about june, July, August, and September (as in Figt1re 4.3a) would show income and employment status

few changes in

between adjacent pairs of those months

but more changes between May (which was covered by an

lier inter­

view) and june; similarly, more changes would be reported

•

·

een

September and October than between other pairs of adjacent months (e.g., june and July). Sea1n effects in SIPP, PSID ( llill,

I

1987), and the Income Development

' We•re grateful to Adriana Silberstein and Monroe Sirken for pointing out previous studies of the seam effect and to Fred Conrad for comments on an earlier draft of this section.

123

Dates and Durations a.

Fnt

Jun

Jul

b. tOO

Feb

...,

All

,...,

Jun

Jut

Aug

.,

Oct

Figure 4.3. (a) Interview schedule for one rotation group in the Survey of In­ come and Prograan Participation (SIPP). (b) Hypothetical forgetting curves for information within SIPP reference periods.

Survey P

(Moore & Kasprzyk, 1984) are quite large. Table 4.3

illustrates this in ter1ns of the percentage change from one month to the next on a number of variables collected in SIPP during 1984 (Young, 1989). The first three columns in this table represent changes between adjacent months in the same reference period. The last column is the percentage change between the final month of the old reference period and the first month of the new one - that is, the transition across the interviewing seam. The percentage change on all variables is greater at this point than for any of the within-reference-period transitions. Re­ ports of Social Security income, for example, change by 12% across the seam but by only 1% or 2 o/o within the reference period. The percentages in the table are calculated over all rotation groups, so the differences cannot be due to seasonal trends. (See Martini, 1989, for similar data on employment status.) The seam e ,.,.

�also persists when we look only

at data that the same respondent provides about him- or herself in both interviews (Moore & Kasprzyk, 1984), so proxy responding is not the source of the difference. What factors are responsible for the seam effect? One obvious possi-

I

•

124

The Psychology of Survey Response

TABLE 4.3 Month-to-Month Changes in Income and Other Variables in the SIPP 1984 Full Panel Longitudinal Research Panel

Between Adjacent Moadu Month 1 10 2 of

Montb2 to 3 of

Month 3 to4 of

Ref.

Ref.

Ref.

Variable

Pera•oct

Period

Period

Month4 of Old to Month 1 of New Ref. Period

Marital status Employment status Personal earnings Total family income Individual Social Security Fa1nily AfDC• receipt p rece1pt Family food

0.3 4.6 5.5 4.9 1.9 0.1 0.2

0.3 5.0 6.3 54 1.6 0.1 0.4

0.3 5.4 6.4 5.1 1.8 0.1 0.2

0.7 10.2 16.3 17.9 12.0 0.3 0.9

•

•

..

AFDC, Aid to Families with Dependent Children.

Source: Young (1989), Table 1.

bility is that memory for the relevant event or quantity decreases across the reference period. Respondents are more likely to remember income they received in the month preceding the interview than income they received four months before the interview. If so, the respondent's data will exhibit a false transition across the seam between adjacent months covered by different interviews. The data in Table 4.3 show a small increasing trend across the first three columns for personal earnings, family income, and employment status, consistent with the idea that respondents remember more changes in the later part of the reference period. (The data in Table 4.3 for changes in marital status, food stamps, and AFDC are too close to zero to be revealing.) The best evidence for forgetting in this context, however, comes from Kalton and Miller's

(1991) study of respondents' memory for the amount of their Social Security benefits. An actual increase in benefits during january 1984 was reported by 68.4% of SIPP respondents interviewed in February; how­ ever, this percentage decreased to 59.6o/o for respondents interviewed in March and to 53.0% for respondents interviewed in April. Thus, reports of the change decrease with tinte, implicating forgetting. Forgetting might be sufficient to explain the seam effect for some variables, but for others it is unlikely to be the whole story. If forgetting were the only cause of the seam effect, then the change at the seam should approximate the total change within the reference period. We

Dates and Durations

I I I

illustrate this effect in Figure 4.3b, which shows hypothetical forgetting curves for the sa11lple rotation group. Respondents report the May data after a one-month interval and the june data after a four-month interval, so forgetting rnay lead to greater underreporting of income in June and thus to a seam effect. However, respondents also report the February figures after four months. On average, then, the February-to-May change (a four-month difference within an interview) should be about the same as the May-to-June change (a one-month difference across interviews). The data in Table 4.3 don't allow us to calculate precisely the total change within the reference period; however, in the case of Social Secu­ rity benefits, it's clear that the change within the reference period is too small to explain the change at the seam. Record checks of SIPP data show relatively srnall effects of forgetting on most variables (Marquis & Moore, 1989). The forgetting explanation that we•ve just considered runs into diffi­ culties because it calls for changes within the reference period as well as changes at the seam. This suggests that seam effects may be due in part to respondents' tendency to minimize change within the reference period (Kalton & Miller, 1991; Young, 1989). Respondents may simply report their current level of income as the value for previous months in an effort to simplify the task and to avoid memory retrieval. This forn1 of response bias is sometimes called a constant wave response. Ross (1988) describes a similar phenomenon - retrospective bias - in autobiographi­ cal memory. l.acking detailed memories about earlier periods of our lives, we may extrapolate from our ct1rrent characteristics (what we eat, how much we drink, what we think about political issues) to our past. If the characteristic is one for which little change is expected (such as political attitudes), we may infer that the past value is identical to the present one. If, on the other hand, the characteristic is one that we expect to change, we may exaggerate the amount of difference between the past and the present in reconstructing the past. For instance, persons who have been through therapy may subsequently exaggerate their pretreat­ ment troubles ( see Ross, 1988). A related possibility is that respondents use their c11rrent level as a starting point for estimating values from earlier months but fail to adjust sufficiently for intervening changes - an example of the anchoring and adjusanent heuristic (Tversky & Kahne­ man, 1974; see Section 5.1.2 for further disa1ssion). .. bias, and anchoring-andConstant wave response, rettos adjustJnent could all produce seam effects since changes will be tempo­ rally displaced into the seam. They will also produce correlated errors

12 S

126

The Psychology of Survey Response for data originating from the same interview, with lower correlations for\ data across the searn (Marquis & Moore, 1989). For exaanple, consider a respondent from the rotation group in Figure 4.3 who

·

receiving

Social Security income in August, and suppose that during the October interview the respondent reports receiving Social Security for june, July, August, and September. Then the change will appear in the

be­ tween May and june because during the jt1ne interview (before receiving the benefit), the respondent reports no Social Sea1rity for May. Kalton and Miller,s (1991) study of the 1984 Social Sec11rity increase provides evidence for displacement of this kind. Young (1989) showed that the seam effect is different in magnitude for different variables (e.g., food statnp recipiency vs. Social Security income). It seems likely that seam effects are more co1runon when mem­ ory retrieval becomes too difficult. The effect varies across questions because retrieval difficulty varies across questions. (Smith & Jobe, 1994, seaan

have developed a related model for survey questions about dietary in­ take.)

4.3.2

Telescoping

We first discussed telescoping- errors in dating events- in Chapter 3, where we considered whether clarifying the boundaries of the refer­

ence period could reduce the frequency of such errors. Here we face the issue of why telescoping occurs in the first place. The most thorough study of telescoping is one of the earliest: Neter and Waksberg's (1964) investigation of respondents' memory for household repairs and modifications. Neter and Waksberg's study was an experimental survey that varied the conditions under which house­ holds were interviewed. It compared unbounded interviews, which sir.n­ ply asked respondents to provide information about jobs during the reference period, with bounded interviews, which provided respondents with a list of jobs from the previous reference period before asking them about the current one. Suppose, for example, that interviews are conducted with a particular respondent in March and again in April. If the April interview is bounded, the interviewer would first infor1n the respondent of the data that he or she bad provided in March and would then inquire about jobs in the current month. If the April inter­ view is unbounded, the interviewer would ask about jobs in the Cllrrent month with no preliminaries. The reference periods themselves were either one month, three months, or six months long. The premise of

Dates and 1)1ara ·ons

127

the study is that bounding reduces telescoping by discouraging respon· dents from reporting jobs they'd already reporting in an earlier inter­ view. Thus, the con,trast between unbounded and bounded interviews I

provides a measure of the impact of telescoping in the usual unbounded conditions. It is worth while to summarize some of Neter and Waksberg's main conclusions. First, the comparison between bounded and unbounded recall for one-month reference periods produced the results in (12):

(12) a. Forward telescoping occurred for both the number of jobs and the total expenditures on them, as indicated by larger reports in the unbounded condition. b. The amount of forward telescoping was greater for larger jobs. c. The amount of telescoping was largely unaffected by uncertainties about completion dates of the job, as indicated by similar effects for do-it-yourself jobs and jobs completed by others.

Figure 4.4 shows some of the data supporting these conclusions. They­ axis of this graph indicates the percentage increase in reported number of jobs for the unbounded reference period compared to the bounded period, that is, 100• (jobs reported in unbounded recall - jobs reported in bounded recall)/ jobs reported in bounded recall. This increase is assurned to reflect net forward telescoping. More expensive jobs exhibit a larger increase than mailer jobs, especially for do-it-yourself jobs. If respondents have a more exact notion of when they finished their own jobs than of when contractots finished their work then precision in knowledge of temporal location does not reduce telescoping. Similar results appeared in the comparison of bounded and unbounded recall for three-month reference periods. Second,

eter and Waksberg estimated the amount of intert•al tele­

scoping that occurred within the three-month period. Respondents had to assign each job to a specific month within this period, and internal telescoping occurred when they assigned one job to the wrong month. To determine the extent of internal telescoping, the investigators com­ pared the monthly data for the bounded three-month period to the data for bounded one-month .periods, making statistical adjustments for pos­ sible effects of forgetting. Their conclusions parallel those in (12): Inter­ nal (forward) telescoping increased the nt1rnber of reported jobs and expenditures in the most recent month at the expense of the earliest month within the reference period. Moreover, the extent of telescoping increased with the size but not the type of job (do-it-yourself vs. other).

I

I

logy of Sun'ey

128

00

..., 11 2

0 ._

� e

C)

=

z

.,

r:::

�CD

20

IQ..

00.-.J tllef

) .4. Percent increase in the

wnber of jobs reported

·

1-mo tb .

-

bounded refer nc

periods compared to bounded 1-month reference periods

(.after

k.sberg, 196,4, Tables 1 and 2). Reprmted with pertmssion

ete.-

from Jounuzl

of the AmerictJn Stlltistical Association.

Copyright C 1964 by the,

American Sta·ti tical A ociation. All rights reserved.

y do resp ndents misreport t e temporal location of events? Two� �explanations ar possible. The first is tha telescoping �;eftects a subjective distonion of the tit11e line, so that (in fo seem closer to the present than

a d telescoping) remote ev

nts,

are. The second possibil·ry is rh t

teleSetor

b tr �ct of th

vent

gr ement about how people r trieve informa­

tion abo�ut these even s. Jon s 111igbt tart with n

though

ago

,

�

e.

lf. but

i ited

and o on)� that

based on these cu

(e.g.

' If I had back pain I prob,ably consulte�d Dr. Alberts � can also fill out '

th information Jon s is using to probe memory. This cycl can continu until Jones ha

much from

�ound the nee�ded facts o�r has extracted

memory a i po ible under the conditio�ns of the interview.

ow suc­

cessful this process, will be depends on the richn ss of th cue th match tween the cue� and the event-as-encoded, and tb

tirtte that's elapsed

ince the event took place among other £ ctors. or many que ri�on , mem.ory won,t uffice to pic o ition ha,t the que tion �eek .

eople u u ly do·n t encode in memory

th calendar date of per. onal event event h ppened

, e pondent

p

perhap� u ing the ort of con traintin Ch pt r 4. Thi proce J n,

' I

.

out the true prop­

·

o if the que tion a k wh n om

bly hav� r· facti

t

econ truct the dat pr ce . th

r

,

e

ut ined

n take into� accoun lan�dmark e ent (e.g.

we ding, birth of her chil , start of her ne

job)

calendar

Cognitive Models and Surveys information (e.g., beginnings and ends of school terms, dates of holi­ days), personal time periods (e.g., when jones lived in Lincoln"' Ne· braska; when Jones was in college), and other facts that have temporal implications. In a similar way, people don't usualJy encode a running tally of the number of occurrences of each type of event they've experi­ enced; so, when Jones is asked how many times he's cons.ulted a doctor, there is no ready-mad,e answer she can retrieve. Jones could try to recall and count individual incidents or she could estimate the number ba ed on typical rates (e.g. that she usually sees a doctor three times a year). Which of these strategies she uses will depend on the perceived magni­ tude of the answer, the le.ngth of the reference period, the availability of the incident information, and the regularity of the event type. Answers to attitude questions, like answers to behavioral questions, usually aren't preformed, waiting for the respondent to retrieve them from long-term memory. If jones has to answer a question about her attitude toward gun control, for example, she is likely to base her answer on a sample of considerations that bear on the question (e.g., con titu· tional rights, dangers associated with gun accidents, criminal activity}, combining them to yield an overall opinion. This sampling proce s helps explain the instability of attitudes ·over time as well as their susc-eptibility to context effect.. Earlier items in the interview can make certain consid­ erations more salient and thus more likely to be included later in the sample for the target item. Of course, not all considerations that come to mind are necessarily incorporated in Jones's answer. Perceived redun­ dancy between items, perceived irrelevance, or warnings. about possible biasing effects can cause Jones to discard �considerations that might oth­ ·erwise have affected her judgment about an issue. Whether the question calls for a judgment about time. frequency, or an attitude issue, Jones is likely to come up with an answer using similar processes. Sometimes

be 11 be able to report an answer

he formed

earlier retrieving an exact date, a tally, or an existing judgment about the issue. Much of the time, though, she,U arrive at an answ;er via some other route. For ·examplet her report may be based on an impression. With time or frequency questions, this impression may reflect how hard it was for her to recall the target event(s); with an attitude question, it may reflect a vague evaluation. Or Jones may make a new judgment derived from general information she recalls at the time th.e qu.estion is asked. With temporal questions, this general information may take the, form of a lifetime period or some other higher-lev·el temporal unit that helps her date the target event; with frequency questions, the general

p

31 7

1

Th P ychology of Surv·ey - e po.nse information may b th

vent' typi al ra . of

questions it tttay in ol e jon '-

retri v

in thinking

1b ur tb

d t il, pi od

lue or i�deologi al pt�e i po­

h he c n

�

judgm n

·

mpute· a · t I, or

infer i�

r pecific cific

dat

n id r ti n ,

·

of h

i� w. R

ardl

mak

Jone i

i call d ·on t

h

he n1ay remem

he c

.h1ch

at an o erall

from whi h she can arriv of th

nswer.

bout her

in�cident from

fr m w ·

natur

ttitude

in lly jone s judgments may be b sed on specific inform tion

ion .

�

. broader

urr�ence·� with

o

lik: ly to u.31:.

on of the e bfoad strate ie to make it. · aUy Jo� e

to� the que rion depe ds

re p ·

the· intern I infor1nation con' tra nt .w r

u

a t

dju t h r

qu

or que tion

ti n

n wer to he p r , iv,ed

nge

·

tr ·nt . If no r�e·:pon e option are pr

h

ab I

I

f

nd t

he m y

eot

other c n-

lso hav' �o� deci�d

e

ct rh re pon e hould b and th

in� ly.)

h

may al o gi e more wei ,h.t to c rt in �e pon e option

'-. nding on their order 'fir t or 1 ( i ual

r

uditory). I

t in a Li t)

he top·�c i

a

an wer 'UP or down or she may refuse to cen'

rin

·

·

·

m r

r

ord-

nd ch nn 1 of pre ent on

m y

h

ion

h d

h r

n .w r e·ntirely. Thi I re pon�L

hen

li ely ro ha.ppen

ro,und h r ns

n i iv

t

How- ble

on ho

11

an-

m 'Y h _v

nd di tribution

pon e , to the perc iv .d m aning of

r

rn ·1

thar demand a nume'fi

actor c n ult ti n

ut

n

�r·....-rw

r triev. d or inferred and the

h

f the .. nterview.

·

n the fir

perc i e

h

r r· -k

I

in

a utitting to orne taboo when an interview· r i pr�e ent than when the ·

survey i betn

I

lf-adtttiniste�e�d : h in a li

c

hi

picrur

traditional �

s

arli r chapt r

I

pr�ofe

he thi�nks rhere i

no d

er in

�

of qu .tion an wering in

urv

y

come

in

ar�ch in cognitive and ocial p ychology. mak

urv Y' ha� I becom Th

nd when

cl ar howev r re ·

· 1 str am

d aim of thi

·

f inv

ve tiga ·on i to

ur ey erro·r and it thu s . ms fair to

t rm . That's me as

arch on c

. we bope o�ur ·

ti e

der tand

k how far it

ct

of

own right.

tigarion in i' ·,

ar from

nd to� r due

c me ·, it own

for th remaind r of this chapt r ( nd th r tnain­

d r of dtis book).

1.2

I

pact on �Coocep ions of I ,\Jr-vey Measur�ement

Perh ps the' most obvious ch nge th CAS th ·

way

ppr

pi

'

-

· -·

movement ha p oduced i

urv y res archers view m asurem nt error . B for

h empiric I in e ti m al character:

tion

robl m wer

of

urv

or

mea

re

th

n w

nt often had a

cata og, d but th· . origin of th

J

Cognitive Model and Survey problems and the relations between the different typ

s,

3 9 ·

of problem wer

often unclear. For e ample,, the 'quality' profiles for several federal su,rveys pro ided a rich and detailed account of the ource of porential ·

errors but made littl�e :effort to develop hypothese about the cau

of

these problems (see e.g. Brooks & Bailar 1978�, and Jabine, 1990). We do not intend thi as criticism; developing theories about the ource of problem . wa not the purpo e of these docut11ents. They do however, ·

reflect the

tate of the art in survey methodology prior to the new

approach. Since the ad ent of the CASM move�ment, re earch r have traced these measurement errors to� the psychological processes of the respondent (e.g., Strack & Martin 1987; Tourangeau 19'84) the inter­ viewer (..··ander et al..,, 1992), and the character of the interaction ( chaef­ fer, 199la·

,

hober, 1999;

·.

uchman & jordan, 1990). Di cu ion, of

mea urement problems in survey now often organize them in term of the mental component

of the response proce

that give ri

to the

problem (e.g., Groves. 1989 Chapter 9), and re earch r have developed ch cklists that ti problems in individual survey questions to the under­ lying cognitive operation (e.g. Le ler & F�or yth, 1995). The cognitiv model of the re pon e proces have provided a ne

paradigm for un­

derstanding- or at least cla sifying- the different type, of mea urement errors 1n surveys. •

1.2.1

S atistical Conceptions of Survey Error

Until recently, the reigning conception of survey errot wa

tati tical

rather than p ychological in character. The new cogniti e models do not ·.

o m,uch contradict the e earlier tatistical model of error a

upplement

them. The statistical models concern the con equences of different type. of urvey error· for the estimat s derived from the survey; the cognitiv models by contrast focus almost

xclusively on the causes of .rrors.

Survey errors� have two main conse,quences for survey estimates. Wh n the errors are systematic, they bia random

the estimates· when the errors ar­

they increase their variance,.

o

. urprisingly, the statistical

model fall into two categorie : variance models and. bia mod·el . The Han en

·

urwitzr-Ber bad Model

...,essler and Kal ·beek (1992) have pointed out, the U. . Censu Bureau ha been the leader in the development of variance model , and no ingle model developed at the Bureau has been more influential than ·

the one propo ed by Hansen, Hurwitz, and Bershad (196 }. The model

I

t1

320

The Psychology of Survey Response assumes that, in effect, in any survey, there is an initial random selection of respondents from a population of potential respondents, followed by a random selection by each respo,ndent of an answer from a distribution of his or her potential answers. Both factors - the sample and the specific response - contribute to the overall variance of the survey esri1natc. The model also allows for the possibility that bias inflates the total error, where bias is conceptualized as the difference between the population mean and the true value for the quantity in question. We give the mathematical details of the model in the appendix to this chapter. This conception of measurement or response errors applies most nat­ urally in situations in which there is a true value. For exa11tple, it is easy to see how it applies when the question asks about the nutttber of doctor visits the respondent made over the last six months or the number of hours he or she worked during the past week. We might expect answers to these questions to be biased downward by forgetting (or upward by telescoping), and we might also expect some variation if we were to readminister the question to the same respondent. It is still possible to apply the model even in situations in which there is no ''platonic'' true score, but the true score is instead defined as the mean across repetitions

of the interview. In such cases - for example, when attitudes are being measured - the model is equivalent to the classical psychometric forrnu­ lation (see Chapter 1 and Biemer & Stokes, 1991). Bias Models Although the Hansen-Hurwitz-Bershad model includes a bias ter1n, its focus is clearly on the variance components of the error. There are, however, a number of statistical models that focus on biases as contrib­ utors to the overall error in survey statistics. One member of this family of models concerns the impact of nonresponse on survey estimates. In the simplest formulation, each member of the survey sample is seen as belonging to one of two subpopulations, or strata: those who would consistently become nonrespondents if they were part of the sample and those who would consistendy become respondents. By definition, data are never obtained for members of the nonrespondent stran1rn. If the unadjusted sample mean is used as an estitnate of the overall population mean, then the nonresponse bias (BN R) will be the product of two factors

- the proportion of the population in the nonrespondent stratum (PNR) and the difference between the means for the two strata (YR and YNa):

I

3 21

Cognitive Models and Surveys More sophisticated formulations express the bias in terttts of the response propensity for each member of the population; this propensity represents the probability that an individual will provide data if he or she is part of the sample. Un,der this stochastic model of nonresponse, the bias in the unadjusted sample mean will depend on the

shortfall in

the sample total due to nonresponse and the covariance of the response propensity and the substantive variable of interest. 3 The shortfall in the total can be eliminated by increasing tbe sample size (or adjusting the sample weights). It is far more difficult to eliminate the portion of the bias that reflects the relationship between response propensities

and the characteristic that he survey is attempting to measure. For

9 that income is related to the likelihood

example, we saw in Chapter

that sacnple members will opt out of the income supplement to the CPS (Moore

et

al.,

1999). It's not easy to compensate for the

bias introduced

by such relationships.

11.2.2

to

Combining the Statistical

and

Cognitive Approaches

r

On the surface, statistical models like the ones st.urunarized here seem at best unconnected with the cognitive models described in the rest of this book. But the great virtue of the statistical error models is their

flexibility. They apply in almost any situation because they Jnake very

or example, it is always possible to

minianal assumptions. as the

ee

an ans er

of the true answer and an error, as in the Hansen et al.

Stirn

model, since the error is

depned as the difference between the reponed

and the true answers. The cognitive models, by contrast, make more substantive assumptions. Their great virtue is that they make stronger predictions about the nature and direction of the errors than the statis­ tical models do. This suggests the possibility of combining statistical models' methods of partitioning error with the cognitive models' ideas 3

Formally, N

Bm(:P)

I (1 - P1)Y1

=

=

where

p, is

the respon

.._t -•

N Cov(p,Y;)- (1 -

--­

-

p)f p is p represents the expected

the response propensity for person j and where propensities

(so that

1

-

the population mean of nonrespon.se rate across

samples).

I

I

I

22

The P ych logy I f Suncy Re ponse ab ut th di

·

ourc.

of these

( ee

rror

rov

1�999� for ·n e fi.nd d

,

i n):'

·

uirchean igh

0

1( 199 ) wor: o

develo men o�f

tr t

th

.d

io

b�

rt igh

.

U i

data

rgu

that t

pie r

·

pon.

�t·sc·c 1 mod 1 that refl r

the

r · te

I

varianc.

illus-

th or ri

·ew pr gram,

tUnat d simpl

uir-

variance - th

resp�ons

ility of the responses across hypothetical J}epe�titions of th

ari .i

- tak

on diff r nt valu s, d

from . elf-report r or pro· i of simple imp , th

p nd·· for a form I d 6nition

1

riance invol es, comparing th data in the

w·th tho e obt ·ned from a reintervie n wers

rror in the

·

the pr

correl t d �c�os

P

.

dure . i 1

.

the

int rth t

urn

·

The a· umpr: o�n

ndenc may not be t n bl when the int· rval between the initial

of inde inr rvi

r

om

usual procedure for e- tim tin

.

pon'

r

inter-

nding on wh th r the d ta

. (See he

v ri nc ) Th

spons

li

n-

I

nd th

·

me p

r interview i

on pr�ovide the da

b r hi

o�r her

ho�rt -

e

both ritne .

-

�or le

·

nd rh

he re p ·ndent m y

n w r a circum tance th t wiU pre

rli r

m m­

r�

_

bly in­

crea e the consi tency of the r sponses and lead to an underestim th

re pon

J�

..

C gniti

t

ampl

n

model

I

u

r

ari nc

e

r

in

thi

dire t r I

by u ing

imp I

.

e ti

�

r p

rts

rror

•

ub tanti

ponden

the nonr

id

lilfe event alend r .

or e

mple

m d

1

are eypi all t

In prin plr other ·

par

=

may r

ould

nt

I

t r

_

el

_

adjusting diru ted f r n

u eful in

rve,

n

nr -

n

.

u

"theor ti aJn mod l, tend

th

I

ub t nrive mod

in order to reduce bi s.

eter and

aJ o p�rovid ak berg

ba

r

964 ), for t"xampl ,

(

f irl

for adju tin

·

) ,

the

,

po

Ii

nd th

urve

r ued that

ropornon retained in m mory

(a )�

' ut

,

r

bout the frequency of home repair: were the product of ou!r factor .. r

of th re pond nt

proportion delib r

f1 ctin (p�), the r I I ft

(t�1):

p� ( l aJ) ( 1 P,) . . t-he 1 m r estun ang -

.

uir

varym

hi h is how

,.-lie ring ·

i · le for

u h techniqu

of th n nr

f th

tual fr�quen ' of relevant r�pairs (

pr

u

rimate

·

err r . It i

e di

tari ti al m

proportion repon d m error due to rete .coping

(

t

ult from the memory lit ratur to provide better cue to re p n

to redu

po

l�

reducing

l o po si. le th t re ults from cognitive model

e ·m·

OU[

i:n

ttempt

h pter.

It i

the

uircheartaigh

�0

metime pi y

develop· u h memory

or

_.

of

�

the length of

erer and

d ition

v It_u

I

�or th e _

ak berg derived th u

infor111 rion a

� ar

pr ctic� of ignorin

f mtr

u ing n

n t

ma

r ference period or other feature e tim

ut the e en

ptible to di tortion by memory error .

orrection for

re Iev nt p rameter

nal

t

of th

( uch t

error inco the

r

v�alu

of th f

p

a)

ords dat ) that i

e tim

larg� sources of error clearly has its dr.a·wb c

d ign

an

m y be rclu t nt to appl

fin I

y. _I t

e

·

. .

rill the

r

not uch

urJent

Models and Surveys

3 23

and compensate for this correlation; in addition, he provides separate esrittlatcs of

the simple response variance for proxies and self­

respondents. , OM

·

·

.

,s model and the belief-sampling model of attitude

responses presented in Chapter 6 both attempt to bring theoretical no­ tions to bear on what appear as random error components in the stan­ dard statistical models. In each case, the new models attempt to pull fixed or

·

tic components out of what had been previously written

off as random. Because estimating these fixed components may require relatively complex alterations to the design of a survey (the addition of reinterviews or the i1nposition of split-ballot designs in which different respondents get the items under different modes of data collection), it seems

\Jnlikely that researchers will routinely apply the new, more theo­

retically based models of survey error. Still, efforts to reconcile the statis­ tical and cognitive approaches to error are likely to continue and likely to yield worthwhile results (e.g., Groves,

11.3

Impact

on

1999).

Survey Practice

If the CASM movement has had a dramatic impact on how researchers conceive survey errors, it has had an equally dramatic impact on their attempts to reduce measurement error. These attempts encompass changes in questionnaire design and methods to assist survey respon­ dents Uobe & Mingay,

11.3.1

Questi ;

1989, 1991).

•

Development and Pretesting

From the outset, the movement to apply cognitive methods to survey

problems has focused on tools for developing questionnaires Uobe & Mingay,

1989; Willis et al., 1991). During the past 15 years, survey

researchers have imported a variety of methods for designing question­ naires, including card sorts, vignettes, focus groups, reaction rime mea­ sures, interaction coding, and cognitive interviewing.

Card Sorts and Vignettes The different techniques have somewhat different aims. The goal of focus groups, card sorts, and vignettes is to explore respondents' cogni­ tive struct1•res for specific do111ains. Focus groups have been popular in

I

•

324

lbe Psychology of Survey . rket research settings for some time, but the federal statistical agencies and their contractors have come to use them only recently. Focus groups are not really a cognitive method, and there is already literarure on their

use

an

extensive

(see Krueger, 1994, for an introduction), so we

will not discuss them here beyond noting that their application in federal surveys seems to reflect increased concerns about the potential mismatch between the concepts that the survey questions presuppose and those that survey respondents actually hold. These same concerns spurred the use of some of the other tools described here. In card sorting, respon­ dents group objects, concepts, or statements into piles based on their apparent similarity. Researchers typically analyze the sort data by means of a statistical clustering procedure (see, e.g., Everitt, 1974) in order to discover the cognitive

structures

underlying the similarity judgments.

Card sons have occasionally aided in developing or improving question­ naires. For example, Brewer, Dull, and jobe (1989) explore people'

such data to

conceptions of chronic medical conditions; their find­

ings provide an empirical basis for improving the conditions checklist in the HIS. (For a more extended discussion of the card sorting technique, see

Brewer & lui, 1995).

Vignettes are short descriptions of hypothetical scenarios, and re-

searchers employ them to t•nderstand how respondents would answer questions about these situations (e.g., Gerber, 1990). One drawb,ack to card sorting is that it is difficult for interviewers to administer in a survey. Vignettes by contrast, can be part of an ordinary interview, thus providing some evidence about varying conceptions of a topic or do­ main. Vignettes were extensively used in redesigning the CPS question­ naire (Martin & Polivka, 1995; see also Esposito, Campanelli, Rothgeb, & Polivka,

1991). In the early stages of that effort, vignettes explored

respondents' conceptions of work by describing various hard-to-classify activities (such as volunteer work or unpaid work in a family business) and helped reveal whether respondents categorized these situations ac­ cording to the CPS definition. A later study examined how changes in the CPS questions about work affected respondents' answers to the vignettes. Table 11.1 displays some of the findings from that study (Martin &. Polivka, 1995, Table 1). The vignettes revealed that respon­ dents' definitions of work sometimes differ markedly from the CPS defi­ nition, with the majority of respondents tnisclassifying two of the scenar­ ios in which the workers (Amy and Sarah} received no pay. A new version of the work question affected classifications, helping in some cases but huning in others.

Models and Surveys

325

TABLE 11.1 Percentage Classifying a Vignette as Work 1991 Old Vasioa Would you report as working last week, not counting wo�k: around the house?

Would you re· pon •• · er as working for pay (or profit) last week?

78

85

Last week, Amy spent 20 hours at home doing the accounting for her husband's business. She did not receive a paycheck.

46

29

Sam spent 2 hoW'S last week painting a friend's house and was given 20 dollars.

61

71

Last week, Sarah cleaned and painted the back room of her house in preparation for .. up an antique shop there.

47

42

Cathy works as a real estate agent for commissions. Last week she showed houses but didn't sign any contract.

89

61

Fred helped his daughter out by taking care of his grandson 2 days last week while the boy's mother worked

13

2

I_ast week, Susan put in 20 hours of volun-

36

4

·

Vignette

Bill attended his coUege classes and got paid one night to tend bar for a fratern.ity

NcwVusioa

last week.

teer service at a local hospital

pie size of approximately 300 per venion and vignette. The first five vipertes meet the Note: CPS definition for work and the final two do not.. Smwce: Data from Martin and Polivka (1995). Copyrisht 0 1995. si of ·. ted with pennison the University of Chicago Press. ·

R

·

Tia••e and Interaction

g

Traditionally, cognitive psychologists have used reaction time mea­ sures to test hypotheses about the mental processes people

use

in carry­

ing out some task (e.g., Sternberg, 1969) or about the way their memory for some on

·

is

·

(e.g., Collins & Quillian, 1969). Studies

reaction rimes to attitude questions have followed this pattern, testing

hypotheses about the process of answering attitude questions, about attitude structure, or both (Fazio et al., 1986; Judd

et

al., 1991; Tour­

angeau, Rasinski, & D'Andrade, 1991). Recently, however, Bassili

I

•

326

The Psychology of Survey Response ( 1996b; Bassili & Scott, 199;6) has extended the use of reaction times to identify poorly worded questions; in addition, the speed with which respondents are able to answer attitude questions rnay predict their susceptibility to context effects. Coding of interactions between respondents and interviewers also serves as a method for identifying problematic survey items (Fowler, 1992}. The proportion of respondents who ask for clarification of the item's meaning or who give unacceptable answers provides an index of the presence and severity of problems. As in the case of response rune, survey methodologists' use of this n1ethod reflects concerns about poten­ tial hitches in the response process, especially about whether respondents interpret the questions as intended. ·•nve •

··

•

•

•

By far the most widely adopted new tool for questionnaire develop-­ ment has been the cognitive interview. As Conrad and Blair ( 1996, p. 1} point out, •'The most tangible result of the dialogue between survey methods research and cognitive psychology is the widespread use of think aloud methods for pretesting questionnaires - so-called cognitive interview s In its pure forrtt, this method requires respondents to repon aloud everything they are thinking as they attempt to answer a survey question; researchers record these reports and analyze them for evidence of misunderstanding and other difficulties. More than ten years earlier, Loftus (1984) had suggested rhat protocol analysis (i.e., analysis of ver­ bal reports about thought processes ) might serve as a useful means for exploring how respondents answer specific survey questions. Shortly thereafter, NCHS set up a cognitive laboratory to carry out cognitive interviewing and conducted a study to evaluate this new method for developing questionnaires. The evaluation pitted a version of a question­ naire on dental health that researchers developed in the traditional way with one they developed using cognitive interviews or other ''cognitive'' methods, such as experimental comparisons of multiple versions of the items (Lessler et aL, 1989). Although it is clear that cognitive interviewing is the direct descendant of the protocol analysis that Herbert Simon and his colleagues invented (Ericsson & Simon, 1980, 1984), the term now has somewhat broader scope, encompassing most of the cognitively inspired procedures we have touched on here. jobe and Mingay ( 1 989), for example, included nine methods in their list of cognitive interviewing techniques: .

9'

I

I

•

Concurrent think-alouds (in which respondents verbalize their thoughts while they answer a question); ·ve think·alouds (in which respondents describe how they

•

arrived at their answers either just after they provide them or at the end of the interview); •

Focus group discussions (in which the respondents take p,art in a semistructured discussion of the topic);

•

Confidence ratings (in which respondents assess their confidence in their answers);

•

Paraphrasing (in which respondents restate the question in their own words);

•

Soning (in which respondents group items based on similarity or rank them on one or more scales);

•

Response latency (in which response rimes are measured);

•

Probes (in which respondents answer follow-up questions designed to reveal their response strategies); and

•

Memory cues (in which the respondents receive various aids to recall).

At 111any survey organizations, the practice of cognitive interviewing appears to encompass a narrower set of activities, including concurrent and

retrospective

protocols, probes designed to identify response strate­

gies, and requests to the respondent to paraphrase items or to define unfarniliar terms (see also Willis & Schechter, 1997). Most of the follow­ up probes (including those eliciting paraphrases and definitions) are scripted ahead of time, but some are generated during the interview. However, it is clear that the method is new, and there are no shared standards for carrying out cognitive interviews (Willis, DeMaio, & Harris-Kojetin, 1999).

. Schemes for Cognitive Interviews Another issue on which there is little consensus is how to code the infor'mation obtained from a cognitive interview. One method is to produce a kind of ''gist'' transcript for each interview, sutrunarizing each respondent's protocol (or answers to probes). Researchers tnay

..

·• er

s11mmarize these transcripts in a report pointing out the problems in each item. An obvious weakness of this and similar procedures is that cognitive interviews become the basis for nonquantitative, essentially impressionist analyses of rhe results of the interviews - hardly a desirable

I

328

The Psychology of Survey . situation. Conrad and Blair (1996, p. 8) note that the success of cogni­ tive psychology is attributable [in part] to the use of rigorous experimental methods that rely survey on objective, quantifiable data. It is ironic, therefore, that the methods community has adapted cognitive psychology as a set of largely impressionist methods. •

.

-

Several researchers, including Conrad and Blair, have attempted to rectify this situation by creating coding schemes that allow more rigor­ ous analysis of the results of cognitive interviews. Conrad and Blair•s scheme groups problems by the cognitive process that gives rise to them - understanding the question, performing the implied task, or forr11atting and reporting the answer. In addition, their coding scheme includes a second component reflecting the aspect of the question responsible for the difficulty. This second component distinguishes five problem lexical problem , which derive from the meaning of the words or their use in the current context; inclusion/exclusion problems, which arise in determining the scope of a ter1n or concept; temporal problems, which involve the boundaries of the reference period or the duration of the activity in question; logical problems, which result &om the logical form or presuppositions of the question; and computational problems, which derive from the capacity of working memory. Quite a few alternative coding schemes have appeared in recent pro­ posals, and we surmnarize them in Table 11.2 (see Blair

et

al., 1991;

Bolton, 1993; Conrad and Blair, 1996; Lessler & Forsyth, 1995; Presser & Blair, 1994; Willis, 1997b). Most of these distinguish trouble spots

associated with each of the major components of the response process that we have discussed in this book: comprehension, memory, judgment, and response formulation. Presser and Blair (1994), however, also code for interviewer difficulties, such as problems in reading the question or recording an answer. Blair et al.'s (1991) scheme is specialized for iden­ tifying the strategies that respondents use to estimate behavioral frequen­ cies, such as the recall-and-count and rate-based estimation strategies that we reviewed in Chapter 5.

11.3.2

Evaluating the New Methods for Questionnaire

Development It might seem obvious that the new methods for developing and I

pretesting survey questions would yield improvements in the questions

I

I

TABLE 1 1.2 Systems for Coding Cognitive Interviews

Basic Categories

Scheme

Other Dimensions/

Additional

Categories

Features

-

Blair, Menon, and Bickart

Strate,gies for frequency ques-

Focus on identifying .re­

(199:1)

ttons

sponse strategy rather

•

Automatic

than question problems

Recall-and-count Rate-based estimation E·numerati�on-based For .Proxy attitude questions Anchoring-and-adjustment Other For recall Search strategies Use or cues Reference period Bolton (1993)

Fourteen categories of verbali-

Automated coding based

zat1on

on presence of key

•

words in verbal protocol Conrad and Blair (1996)

Three response components

Five .problem types

Understanding

Lexical

Task performanc

Temporal

Response forntatting

Logical Computational Omission/inclusion

(continued) ::::T

w N \.0

-

-

..

•

....

��

� -•

•

blic Opinion Quarterly, 52,351-364. McCloskey, M., Wible, C., & Cohen, N. (1988). Is there a special flashbulb memory mechanism? journal of Experimmtal Psychology: General, 1 J 7, 171..

..

181.

McGill, A. (1989). Context effects in judgments of causation. }o11rnal of Person­ ality and Social Psychology, 57,189-200. McGuire, W. J. (1960). A syllogistic analysis of cognitive relationships. ln M. Rosenberg, C. Hovland, W. McGuire, R. Abelson, & J. Brehm (Eds.), AUitMJ� organiZiltion and change (pp. 65-111). New Haven, CT: Yale University Press. McMullen, M. (1997). ve contrast and assimilation in counterfactual thinking. JourntJI of Expni�ntal Social Psychology. 33, 77-100. McQuceJt, D. V. (1989). Comparison of results of personal interview and tele­ phone surveys of behavior related to risk of AIDS: Advantages of telephone techniques. In Conference Proceedings: Health Survey Research Methods (pp. 247-252) (DHHS Pub. No. (PHS) 89-3447). Washington, D.C.: U.S. Department of Health and Human Services. Means, B., � Loftus, E. (1991). When personal history repeats itself: Decom­ posing memories for recurring events. Applied Cognit;ve Psychology. 5, 297·

318.

Means, B.,Nigam, A., Zarrow, M., Loftus, E., 8c Donaldson, M. (1989). Auto­ biographical memory for health-related events. Vital t�nd Hulth St4tistia, Series 6,No. 2 (DHHS Pub. No. (PHS) 89-1077). Washington, DC: U.S. Government Printing Office. Means, B., Swan, G. E., Jobe, j. B., & Esposito, J. L. (1991). An alternative approach to obtaining personal history data. In P .. Bietller, R. Groves. L. errors in sl4rveys Lyberg, N. Mathiowea, &c S. Sudman (Eds.), M (pp. 167-184). New York: Wiley. Means, B., Swan, G. E., Jobe, j. B., & Esposito, J. L. (1994). The effects of estianation strategies on the accuracy of respondents' reports of cigarette smoking. InN. Schwarz&: S. Sudman (Eds.), Autobiographia�l m6'''ory and th� v"lidity of ret:tospeaive reports (pp. 107-119). Berlin: Springer-Verlag. Menon, G. (1993). The effects of accessibility of information on judgments of behavioral frequencies. JourMI ofConsu�r Rese11rch, 20,431-460. Menon, G. (1996). Are the parts better than the whole? The effects of tkcompositioMI on jNdgPM�•ts of frequent behaviors. Paper presented at ·

3 64

References the Conference on the Science of Self-Repon, Bethesda, MD, November 7, 1996. Menon, G., Raghubir, P., & N. (1995). Behavioral frequency judgments: An accessibility-diagnosticity • jour1tal of Consume�' Research, 22, 212-228. Mieczkowski� T., Barzelay, D., Gropper, B., & Wish, E. (1991). Concordance of three measures of cocaine use in an arrestee population: Hair, urine, and self-report. Journal of Psychoactive Drugs. 23,241-249. Mieczkowski, T., & Newel, R. (1997). Patterns of concordance between hair assays and urinalysis for cocaine: Longitudinal analysis of probationers in Pinellas County, Florida. In L. Harrison & A. Hughes (Eds.), Tht validity of self-reported drug use: Improving the accuracy of SMrvey estimates (pp. 161199). Rockville MD: National Institute on Drug Abuse. Millar, M.G., & Tesser, A. (1986). Thougbt·induced attitude change: The ef­ fects of schema structure and coJnanitment. ]ou,,•al of PnsoMlity and Social Psychology, 51, 259-269. to official Miller, P. B., &: Groves, R. M. (1985) Matching survey records: An exploration of validity in victimiza tion reporting. PNblic Opinion Quarterly, 49, 366-380. Mingay, D. J., ShcveH, S. K., Bradburn,N. M., & C. (1994). Self and proxy reports of everyday events. InN. Schwarz & S. Sudman (Eds.), Auto­ biographical menrory and validity of retrosp1aive r�orts (pp. 225-250). New ·

York: Springer-Verlag. Moore, J. C. (1988). Self-proxy response status and survey response quality. journal of OfPcial Statistics. 4, 155-172. Moore, J. C., &: Kaspnyk, D. (1984). Month-to-month recipiency turnover in the ISDP. Proceedings of the Seaion on Survey Research Methods, Amsrican Statistical Association (pp. 210-215). Alexandria, VA: American Statistical Association. Moore, J. C., Stinson, L. L., & Welniak, E. J. (1999). Income reporting in sur­ ·ve issues and measure1nent error. In M.G. Sirken, D. J. Herr­ veys: mann, S. Schechter, N. Schwarz, J. M. Tanur, & R. Tourangeau (Eds.), Cog­ nition and Survey Research. ew York: Wiley. Morris, M. (1993). Telling tails explain the discrepancy in sexual parb1er r� pons. Nature, 365, 437 . Morris, M. W., & Murphy,G. L. (1990). Converging operations on a basic level in event taxonomies. Memory & Cognition, 18,407-418. Mosher, W. D., & Duffer, A. P., Jr. (1994). Experiments in survey d4ta colkc­ tion: The National Survey of family Growth pretest. Paper presented at the meeting of the Population Association of America, May, 1994, Miami, FL. Moss, L., &Goldstein, H. (1979). The recall method;, soci4l surv�s. London: University of London Institute of Education. Mott, F. (1985). Evaluation of fertility data and preliminary liMlytic results from the 1983 survey of the Natio1111l Longitlldinal Stm�eys of Work Experi­ ence of Youth. A report to theNational Institute of Child Health and Human Development by the Center for Human Resources Research, january 1985. Moxey, L. M., & Sanford, A. J. (1993). Communicating q1111ntitks. Hillsdale, NJ: Erlbaum. .�

I

365 MueUer, J. (1973). War, l'rssid•fw, 4nd public opinio,. New York: Wiley. Murray, D., O'Connell, C., Schmid, L., & Perry, C. (1987). The validity of smoking self-reports by adolescents: A reexamination of the pipeline procedure. Addictive Behaviors, 12, 7-15. Myers, S. L. (1980). Why are crimes underreported? What is the crime rate? Uy matter? Social Science Quarterly, 61, 23-42. Does it yan, S., & Krosnick, j. (1996). Education moderates some response effects N in attitude measurement. Public Opinion Q114rkrly, 60, 58-88. Nass, C., Fogg, B. j., & Moon, Y. (1996). Can computers be tea1ruuates? Inter­ ·

national ]ourMI of Hu1•1an-Computer Studies, 45, 669-678. Nass, C., Moon, Y., & Green, N. (1997). Are machines gender neutral? Gender­ stereotypic responses to computers with voices. journal of Applied Social 76. Psychology, 27, 8 N G., Sirken, M., Willi� G., & Esposito, j. (1990). Laboratory experi· m41'Jts on th� cognitive aspects of sensitive questions. Paper presen ted at the International Conference on Measurement Error in Surveys, Tucson, AZ, November, 1990. Neisser, U., & Harsch, N. (1992). Phantom flashbulbs: False recollections of bearing the news about ChaUenger.ln E. Winograd 8c U. eisser (Eds.), Affect and aca1racy in recall (pp. 9-31). Cambridge: Cambridge University Press. Neter, J., & Waksberg, j. (1964). A study of response errors in expendinares data from household interviews. journal of th� American StlltistiCIJI Associa­ tion, 59, 17-55. Newell, A. (1973). You can•t play 20 questions with nature and win. In W. G. Chase (Ed.), Visual information processing (pp. 283-308). New York: Aca­ demic Press. Newstead, S. E. (1988). Quantifiers as · concepts. InT. zetenyi (Ed.), Fuzzy sets m psychology (pp. 51-72). Amsterdam: Elsevier. Newtson, D. (1973). Attribution and the unit of perception of ongoing behavior.

Journal of P�rsonality and Social PsychologyJ 28, 28-38. Nicholls, W. L., 0, Baker, R. P., & Marrin, J. (1997). The effect of new data collection technologies on survey data quality. In L. Lyberg, P. Biemer, M. Collins, E. deLeeuw, C. Dippo, N. Schwarz, & D. Trewin (Eds.), Survey arul procus quality (pp. 221-248). New York: Wiley. Nisbett, R. E., & Wilson, T. D. (t9n). TeUing more than we can know: Verbal reports on mental processes. Psychological Review, 84, 231-259. Norma n, D. A. (1990). Th� tksign of everydtry things. New York: Doubleday. Nottenburg, G., & Shoben, E. J. (1980). Scripts as linear orders. joumal of '&perirruntal SocUJI Psychology, 16, 329-347. O'Muircheartaigb, C. (1991). Simple response variance: Estimation and deter­ minants. In P. Biemer, R. Groves, L. Lyberg, N. Mathiowetz, & S. Sudman errors in surveys (pp. 551-574). New York: Wiley. (Eds.), O'Reilly, j., Hubbard, M., I.essler, j., Biemer, P., &: Tu1·ner, C. (1994). Audio and video computer assisted self-interviewing: Preliltlinary tests of new tech­ nology for data collection. Jounral of O(pdal Statistics, 10, 197-214. Ornstein, R. E. (1969). On the of time. New York: Penguin. Osherson, D. N., Smith, E. E., & Sha6r, E. B. (1986). Some origins of belief. ·

·

CopitiOPI, 24, 197-224.

I

•

366

References Ostrom, T. M., & Upshaw, H. L. (1968). Psychological perspective and attitude change. In A. C. Greenwald, T. C. Brock, & T. M. Ostrom (Ed.s.), Psycholog­ ical foundations of anitudes (pp. 65-111 ) . New York: Academic Press. Ottati, V., Riggle, E., Wyer, R , Schwarz, N., 8c Kuklinski, j. (1989). Cognitive and affective bases of opinion survey responses. }our11t11 of Personality and ..

Soci4/ Psychology, 57, 404 415. Padian, N. S. (1990). Sexual histories of heterosexual couples with one HlV infected partner. Ameria�n Journal of PMblic Health, 80, 990-991. Panel on Privacy and Confidentiality as Facton in Survey Response. (1979). DC: Privacy and corifUkr•tiality as factors in SurtJeY response. W ·

·

National Academy of Sciences. Parducci, A. (1965). Category judgment: A range-frequency model. Psychologi­

cal Review, 72, 407-418. •• ·s. In E. Carfrequency Parducci, A. (1974). Contextual effects: A terette & M. F riedn1an (Eels.), of perception: Psychophysiul judg.. . ment and musure��nt. (Vol. U, pp. 127-141). New York: Academic Patrick, D. L., Cheadle, A., Thompson, D. C., Diehr, P., Koepsell, T., & Kinne, A review and metaS. (1994). The validity of self-reported ·

·

Amerit41J }oumtd of Publie Health, 84, 1086-1093. Payne, J. W., Bettanan, J. R., 8c Johnson, E. J. (1993). The adaptive decision maker. Cambridge: Cambridge University Press. Pe lham, B. W., T. T., & Myaskovsky, L. (1994). The easy path &om many to much: The numerosity heuristic. Cognitive Psychologylt 26, 103-133. . In Pepper, S. (1981). Proble1ns in the quantification of frequency D. W. Fiske (Ed.), New dir�ctions for �thodology of soci11l and bsh11vioral sciences (Vol. 9, pp. 25-41). San Francisco: jossey·Bass. Petty, R. E., & Cacioppo, J. T. (1981). Attitudes and persuasion: CWsical and contemporary approaches. Dubuque, lA: Brown. Petty, R. E., & Cacioppo, J. T. (1984). The effects of involvement on responses to argu1nent quality and quantity: Central and peripheral routes to persuasion. ·

.

Journal of Personality and Social Psychology, 46, 69-81. Petty, R.. E., &: Cacioppo, j.

and peripheral routes to .

(1986).

·

·

·

and persuasion: Central

char�ge. New York: Springer-Verlag.

Petty, R.. E., &: Wegener, D. T. (1993) Flexible correction processes in social judgment: Correcting for context-induced conuast. ]Otm'IQI of &perit�a.ntal

Social Psychology, 29, 136-165. Phipps. P., &: Tupek, A. (1990). Assessing musurement error s ;, a touchtone recognition survey. Paper presented at the International Conference on Mea­ surement Errors in Surveys, Novctnbu 1990, Tucson, AZ. Pillemer, D. B., Goldsmith, L. R., Panter, � T., &: White, S. H. (1988). Very long-ter111 memories of the first year in college. ]ounral of ExperitMHtal Psy­

chology: Learning, Memory, and Cogniti011, 14, 709-715. Pillemer, D. B., Kreosky, L., Kleinman, S. N., Goldsmith, L. R., &: White, S. H. in rives: Evidence &om oral histories of the first (1991 ) . Chapters in college. Joumlll of N11rtive ra and Life Hutory, 1, 3-14. Pillemer, D. B., Rhinehart, E. D., & White, S. H. (1986). Mensory of life tions: The first year in college. Human Le4ming, S, 109-123. •

-

I

367

Poulton, E. C. (1989) BUu in quantifying judgments. Hillsdale, NJ: Fzlbaum. tion of experience. Poynter, W. D. (1983). Duration judgment and the Msmory & Copition, J 1, 77-82. Pratt, J. W., Raiffa, H., & Schlaifer, R. (1964). The foundations of decision under uncertainty: An elementary exposition. journal of th� �c:a, Statis· ·

tic.al Association, S9, 353-375. Presser, S. (1990). Can changes in context reduce vote overreporting in surveys? P11blic Opinion Qwlrterly, 54, 586-593. r, S., & Blair, J. (1994). Survey pretesting: Do different methods produce different results? In P. V. Marsden (Ed.), SociologiC4I methodology (Vol. 24, pp. 73-104). Beverly Hills, CA: Sage. Presser, S., Blair, j., & Triplett, T. (1992) Survey sponsorship, response rates, and response effects. Social Scimc� Quarurly, 73, 699-702. Presser, S., & Stinson, L. (1998). Data coUecrion mode and social desirability bias in self-reponed religious attendance. American Sociological Review. 63, ..

137-145.

, S., & Zhao, S. (1992). Attributes of questions and interviewers as , 56, determinants of interviewing perforanance. Public Opinion 236-240. Priester, J., & Petty, R. (1996). The gradual threshold model of ambivalence: Relating the positive and negative bases of attitudes to subjective ambivalence. 9. Journal of P and Social Psychology, 71, 431 Quadrel, M., Fischhoff, B., &: Davis, W. (1993). t (in)vulnerabiliryii Americ.an Psychologist, 48, 102-116. Raden, D. (1985). Strength-related attitude dimensions. Social Psychology Quar· terly� 48, 312-330. Radford, A. (1997). Syntactic theory and the strNchlre of English: A minin•alist ·bridge University Press. approach. Cambridge; England: , F. (1931). Truth and probability. In R. B. Braithwaite (Ed.), The foun­ datio7fS of milthematia tmd other logia�l eSSQYI (pp. 156-198). London: Routledge and Kepn Paul. Rasinski, K. A. (1989). The effect of question wording on support for government spending. P11blic Opinion Quarterly, 53, 388-394. Rasinski, K. A., Baldwin, A. K., Willis, G. B., 8c jobe, J. B. (1994). Risk and loss pnaptions associated with Sllrlley reporting of sensitive behlllo li rs. Paper .....ted ar the annual meeti ng of the American Statistical Association, Au­ gust 1994, Toronto, Canada. R.asinski, K. A., Mingay, D., & Bradburn, N. M. (1994). Do respondents really "mark all that applyn on self-administered questions? Public Opinion QUIIr· tnl,, 58, 400-408. Rasinski, K. A., & Tourangeau, R (1991). Psychological aspects of j .: about the economy. Politic.al Psychology, 12, 27-40. Reder, L. (1987). Strategy selection in question answering. Cognitive Psychol­ ogy, 19, 90-138. Reeves, B., & Nass, C. (1996). Th• mt�dia equation: How people trut ers, ulevision, and new wwdi42 like r�al peopk and places. Cambridge: CSLI and Cambridge Univenity Press. ·. .

·

·

·

..

·-

I

3 68

References Reichenbach, H. (1947). Elements of symbolic logic. New York: Free Press. Reiser, B. j., Black, J. B., & Abelson, R. P. (1985). Knowledge structures in the organization and retrieval of autobiographical memories. Cognitive Psychol­ ogy, 17, 89-137. Rifkin, A. {1985). Evidence for a basic level in event taxonon1ies. MeHamy & Cognition, 13, 538-556.

Rips, L. J. (1995). The current status of research on concept com ·on. Mind & Language, 10, 72-104. Rips, L. j., Shoben, E. j., & Smith, E. E. (1973). distance and the verification of nelations . journal of Verbal Learning and Verbal Behavior, 12, l-20. Robinson, j. A. (1986). Temporal reference systems and autobiographical mem­ ory.. In D. C. Rubin (Ed.), Autobiographical PMmory (pp. 159-188). Cam­ ·

·

.

·

.·

bridge: Cambridge University Press. Robinson, J. A. (1992). First experience memories: Contexts and functions in personal histories. In M.A. Conway, D. C. Rubin, H. Spinnler, & W. A. Wagenaar (Eds.), Theoretical perspectives on autobiographical tMntory (pp. 223-240). Dordrecht, the Netherlands: Kluwer. Roese, N. (1997). Counterfactual thinking. Psychological B.Jietm, 121, 133-

148. Rokeach, M., & Baii-Rokeach, S. (1989). Stability and change inAmerican value priorities, 1968-1981. Amsrican Psychologist, 44, 775-784. Rosch. E. H. (1973). On the intertlal structure of perceptual and semantic cate­ go ries In T. E. Moore (Ed.), Cognitive development and the acquisition of language (pp. 111-144). New York: Academic Press. Rosch, E. (1975}. Cognitive reference points. Cognitive Psychology, 7, 532-541. Rosch, E. H. (1978). Principles of categorization. In E. Rosch 8c B. B. Uoyd .

(Eds.), Principles of categorization (pp. 27-48). Hillsdale, NJ: Erlbaum. Ross, M. (1988). The relation of implicit theories to the consr.r·ucti on of perso nal histories. Psychologia�l Reviftll, 96, 341-357. Ross, M., & icoly, F. (1979). Egocet1tric biases in availability and attribution. Journal of Personality and Social Psychology, 37, 322-336. Rubin, D. C. (1982}. On the retention function for autobiographical memory. journal of Verbal Learning and Verbal B•havior, 19, 21-38. Rubin, D. C., lk Baddeley,A. D. (1989). Telescoping is not rime compression:A model of the dating of autobiographical events. Memory & Cognition, 17, 653-661. Rubin, D. C., & Wetzel, A. E. (1996). One hundred of forgetting: A quantitative description of retention. Psychological Review� 103, 734-760. Rubin, D. C., Wetzler, S. E., & Nebes, R. D. (1986). Autobiographical memory across the lifespan. In D. C. Rubin (Ed.), Autobiographit:��l m8mory (pp. 202221). Cambridge, England: Cambridge Uoivenity Press. Sadock, J. M. (1977). Truth and approximations. In K. Whistler, R. D., Van Valin, Jr., C.. Chiarello, J. j. jaeger, M. Petruck, H. Thompson, R. javkin, & A. Woodbury ( Ed s.} , Proa�di,gs of the third tmnual meeting of the Bnkeley ·39). Departmen t of Linguistics, University of Linguistics Society (pp. 43 California, Berkeley.

I

369 Sadock, j. M. (1981). Almost. In P. Cole (Ed.), Radical pragmatics (pp. 257272).New York: Academic Press. Sanbonmatsu, D., 8c Fazio, R. (1990). The role of attitudes in memory-based decision-making. Journal of Personality and Social Psychology, 59, 614622. Sander,j., Conrad, F.,Mullen, P., & Herr1nann, D. (1992). · ·vc modeling of the survey interview. 1992 Prouedings of the Seaicm on S1n11ey Research · VA: American Statistical Association. Methods (pp. 818-823}. Saris, W., &: Pijper, M. (1986). Co�mputer assisted interviewing using home computers. Europemr Research, 14,1 150. Schab,F. R.,&Crowder,R.G. (1989). Accuracy of temporal coding: Auditory& Cognition, 17,384-397. visual comparisons. M Scbacter, D. L. (1987). Implicit memory: History and current status. joM�rnal of Experimental Psychology: Uaming, Memory, and Cognition, 13,501-518. Schaeffer,N.C. (1980). Evaluating race-of-interviewer effects in a national sur­ vey. Sociologiad Methods and R�search, 8, 400-419. Schaeffer,N.C. (199la). Hardly ever or constantly? Group comparisons using vague quantifien. P14blic Opinion Quarterly, 55,395-423. Schaeffer,N.C. (t991b). Conversation with a purpose - or c rion? Interaction in the standardized interview. In P. P. Biemer, R. M. Groves, L. E. Lyberg, N .. A. Mathio''*--etz,& S. (Eds.), Measurement error m 1urvrys (pp. 367-391).New York: Wiley. Schaeffer,N.C. (1994). Erron of experience: Response e�rors in reports about child support and their implications for questionnaire design. InN. Schwarz & S. Sudman (Eds.), Autobiographical memory and the validity of r�trospec­ tivl reports (pp. 141-160). Berlin: Springer-Verlag. Schaeffer,N.C. (in press). Asking questions about threatening topics: A selective overview. In A. Stone, j. Turkkan,C. Bachrach, V. Cain, j. Jobe, & H. K urtzman (Eds.), The science of ulf-report: ImpliaJtions few ruearch and �Wactiu. Mahwah,NJ: Erlbaum. Schaeffer,N.C., & Barker, K. (1995). Issues in using bipolar response catego· rks: Numeric labels and th� middle category. Paper presented at the annual meeting of the American Association for Public Opinion Ft. Lauder­ FL, May 23,1995. Schaeffer, N.C., & Bradburn, N. M. (1989). Respondent behavior in tude estimation. Journal of the Am erican Statistical Association, 84, 402·

· ..

413. _..

fer,

R.

(1992). Reulling

New York: Basic Books. Schank, R.C. (1975). Holland. Schank, R.C. (1982).

.·

11

life: Narration and dialogtu! in psychoanalysis.

·tual information proassing. Amsterdam: Northic

memory.

Cambridge: Cambridge University

Press. Schank,R.C.,&: Abelson,R. P. (1977). Scripts, plans, goals, and untkrstllnding. Hillsdale,NJ: Erlbauan. Schober, M. (1999). Making sense of questions: An interactional approach. In M.G. Sirken, D. j .. Herrmann, S. Schechter,N. Schwarz,j. M. Tanur, 8c R.

I

3 70

References Tourangeau (Eds.), Cognitio" arul SJI1W1 research (pp. n-93). New York: Wiley. Schober, M. F., & Clark, H. H. (1989). Understa nding by addressees and over­ hearers. Cognitiv� Psychology, 21, 211-232. Schober, M. F., & Conrad, F. G. (1997). Does conversational interviewing reduce survey measurement error? Public OpitUon 60, 576-602. Schober, S., Caces, M. F., "t, M., 8c Bran� L. (1992). Effeccs of mode .. .. of administration on reporting of drug usc in the National Sur· vey. In C. Turner, J. Irssler. & J. Gfroerer (Ecls.Jt Survey tMtJSIIf'mtent of ·

,

drug use: Methodological studi�s (pp. 267-276). Rockville, MD: National Institute on Drug Abuse. D. J. (1992). There is more to episodic meznory Schooler, J. W., & than just episodes. In M. A. Conway, D. C. Rubin, H. Spinnler, 8c W. A. Wagenaar (Eds.), Tbeoretia�l perspectives on autobiographical memory (pp. 241-262). Dordrecht, the Netherlands: Kluwer. Schuman, H. {1972). Attitudes vs. actions versus attitudes vs. attitudes. Public Opinio" Quarterly, 36, 347-354. Schuman, H. (1992). Context effects: State of the art/state of the past. In N. Schwarz & S. Sudntan (Eds.), Context �ffects in social and psychologiazl

res�arch (pp. 35-47). New York: Springer-Verlag. Schuman, H., & Converse, j. (1971 ). The effects of black and white interviewers on white respondents in 1968. Publie Opinion QIUirterly, JS, 8. Schuman, H., & ludwig. j. (1983). The norm of evenhandedness in surveys as in life. AmeriC4n SociologietJI Review, 48, 112-120. sJWVeys: Schuman, H.� & Presser, S. (1981). Questions and answers in Experiments in 'f'"Siion fo'"'� wording, arul context. New York: Academic Press. Schwarz, N. (1990). Assessing frequency reports of mundane behavion: Contri­ butions of cognitive psychology to questionnaire construction. In C. Hendrick & M. Clark (Eds.), Review of personality arul scxial psychology (Vol. 11, pp. 98-119). Beverly Hills, CA: Sage. Schwarz, N. ( 1996). Cognition and communieation: Jlldgmentlll biaus, reset�reb methods, and the logic of convrrsation. Mahwah, NJ: Erlban1n .. Schwarz, N., & Bienas, J. (1990). What mediates the impact of response alter­ natives on frequency reports of mundane behaviors? Applied Cognitive Psy­ chology, 4, 61-72. Schwarz, N., & Bless, H. (1992a). Constructing reality and its alternatives: Assimilation and contrasts effects in social judgment. In L. L. Martin & A. Te ser (Eds.), The construction of socilll jlldgment (pp. 217-245). Hillsdale, NJ: Erlbaum. Schwarz, N., & Bless, H. (1992b). Scandals and public trust in politicians: Assimilation and contrast effects. P�rsonality and Social Psychology Bulletin, 18, 574-579. Schwarz, ... Bless, H., & Bohner, G. (1991). Mood and persuasion: Affective states influence the procession of persuasive communications. Advanas in Expnimmtal Soeial Psychology� 24, 161-199. Schwarz, N., Bless, H., Strack, F., Klulllpp, G., Rittenauer-Schatka, H., & Si-

I

R

371

mons, A. (1991). Ease of retrieval as information: Another look at the availa­ bility heuristic. journal of Personality and Social Psychology, 61,195-202. Schwarz, N., & Clore, G. L. (1983). Mood, misattribution, and judgments of well-being: Informa tive and directive functions of affective states. Journal of Perso1111lity and Social Psychology, 45,513-523. N., & Hippler, H.-j. (1987). What response scales may teD your Sch .,............_: Information functions of response alternatives. In H.-j. Hippler, N. Schwarz, & S. Sudt11an (Eds), Social m{on1tation proassing t�nd survey methodology (pp. 163-178). New York: Springer-Verlag. Schwan, N., & Hippler, H..-J. (1995). Subsequent questions may influence an­ swers to preceding questions in mail surveys. Public Opi,ion Qalartnly, 59, 93-91. Schwarz, N., Hippler, H.-J., Deutsch, B., & Strack, F. (1985). Response catego­ ries: Effects on behavioral reports and comparative judgments. Public Opinion , 49,388-395. Schwarz., N., Hippler, H., & Noelle-Neumann, E. (1991). A cognitive model of response-order effects in survey measurement. In N. Schwarz & S. Sudman (Eds.), Contut effects in social and psychological research (pp. 187-201). New York: Springer-Verlag. Schwar� N., Knauper, B., Hippler, H.-J., Noelle- eumann, E., &: Clark, F. (1991). Rating scales: Numeric values tnay change the meaning of scale labels. Public Opinion Quarterly, 55,618-630. Schwarz, N., Sttack, F., & Mai, H. (1991). Assimilation and conuast effects in part-whole question sequences: A conversational logic analysis. Public Opin­ ion Quarurly, 55,3-23. Schwarz, N., &: Sudrnan, S. (1992). Context effects in social and psychologiazl research. New York: Springer-Verlag. Searle, J. (1969). Spe�ch acts. Cambridge: Cambridge University Press. Sears, D. 0. (1983). The person·positivity bias. Journal of PersoNJlity and Social Psychology, 44, 233-250. Sheatsley, P. (1983). Questionnaire construction and item writing. In P. Rossi, J. Wright, & A. Anderson (Eds.), Handbook of survey research (pp. 195-230). New York: Academic Press. Sheingold, K., 6c Tenney, Y. J. (1982). Memory for a salient childhood event. In U. Neisser (Ed.), Memory observed (pp. 201-212). New York: Freeman. Shimizu, 1., & Bonham, G. (1978). Randomized response technique in a national survey. }oun1al of the A1•terican Statistical Association, 73, 35-39. reb, Northwestern University. tion Shun1, M. (1997). Unpublished · Shurn, M. (1998). The role of temporal landmarks in autobiographical memory processes. Psychologiazl Bulletin, 124,42 2. Shyrock, H. S., Siegel, j. S., & Stockwell, E. G. (1976). The �thods aHd n•ate ­ rials of demography (condensed ed.). San Diego, CA: Acadeanic Press. Siegel, A. W., Goldsmith, L. T., & Madson, C. R. (1982). SkiD in estitnation problems of extent and numerosity. }our;nal for Research in Matbematia EdMcation, 13,211-232.. Sikkel, D. (1985). Models for memory effects. Journal of tht American Statutiazl Association, 80, 835-841. ·

·

·

·.

I

3 72

References Silver, B. D., Abramson, P.R., & Anderson, B. A. (1986). The presence of others and overreponing of voting in American national elections. Publie Opinion Quarterly, 50, 228-239. Simon, H. (1957). Models of man.New York: Wiley. Simon, H. A., & Feigenbaum, E. A. (1964). Effects of similarity, familiarization, and meaningfulness in verbal learrting. Journal of Verbal Learning and Verbal Behavior, 3, 385-396. Singer, E., Hippler, H., & Schwarz, N. (1992). Confidentiality assurances in surveys: Reassurance or threat. International jm�r'tlal of P11blic Opinion Re· ut�rch. 4, 256-268. Singer, E., Mathiowetz,N., & Couper, M. (1993). The impact of privacy and confidentiality concerns on survey participation: The case of the 1990 U.S. census. Public Opinion Quarurly, 57, 465-482. and Singer, E., von Thul'll, D., & Miller, E. (1995). Confidentiality response: A quantitative review of the experimental literature. Public Opinion QJUJrterly, 59, 66-77. Singer, M. (1985). Mental processes of question answering. In A. C. Graesser & J. B. Black (Eds.), The psychology of questions (pp. 121-156). Hillsdale,NJ: Erlbaum. Skowronski, j. J., Betz, A. L., Thompson, C. P., & Shanno� L. (1991). Social memory in everyday life: The recall of self-events and other-events. }ounu�l of Personality and Social Psychology, 60, 831-843. Skowronski, j. j., Betz, A L., Thompson, C. P., & Walker, W.R. (1994). The impact of differing memory domains on event-dating processes in self and proxy reports. InN. Schwarz&: S. Sudman (Eds.), Autobiographia�l m41fJory ttnd the validity of retrospective reports (pp. 217-234).New York: Springer· Verlag. Smith, A. F. (1991). Cognitive processes in long-term dietary recall. Vital and Health Statistics, Series 6, No. 4 (DHHS Pub. o. PHS 92-1079). Washing­ ton, DC: U.S.Goverrunent Printing Office. Smith, A. F., & jobe, j. B. (1994). Validity of reports of long-tertn dietary memories: Data and a model. InN. Schwarz & S. Sudman (Eels.), AMtobio­ graphical memory and the validity of retrospective reports (pp. 121-140). Berlin: Springer-Verlag. Smith, A. F., jobe, j. 8., & Mingay, D. (1991).Retrieval from memory of dietary information. Applied Cognitive Psychology, 5, 269-296. Smith� E.R. (1999). ew connectionist models of mental representation: Impli­ S. Schechter, N. cations for survey research. In M.G. Sirken, D. J. Schwarz, j. M. Tanur, & R. Tourangeau (Eds.), Copition and survey re· search (pp. 251-266). ew York: Wiley. Smith, T. W. {1983). An experimental comparison between clustered and scattered scale items. Social Psychology QJUtrterly, 46, 163-168. Smith, T. W. (1984a). Non-attitudes: A review and evaluation. In C. F. Turner & E. Manin (Eds.), Surveying subjective phenonrena (Vol. 2, pp. 215-255). cw York: RusseU Sage Foundation. Smith, T. W. (1984b), A comparison of telephone and personal interviewing.

•

..

·

I

•

References GSS Methodological Report No. 28. Chicago: National Opinion Research Center. Smith, T. W. (1986). Condltioru�l order effects. GSS Methodological ReportNo. 20. Chicago:National Opinion Research Center. Smith, T. W. (1987). That whlch we call welfare by any other name would smeU sweeter: An analysis of the impact of question wording on response patterns. Public Opinion Quarterly, 51, 75-83. Smith, T. W. (1988). Ballot position: An analysis of co,uxt effsct5 r1lated to rotation design. GSS Methodological Report o. SS. Chicago:National Opin­ ion Research Center. Smith, T. W. (1992a). Thoughts on the nature of context effects. InN. Schwarz & S. (Eds.), Context effects m social and psychologiCtZI research (pp. 163-184).New York: Springer-Verlag. Smith, T. VI. (1992b). Discrepancies between men and women in reporting number of sexual partners: A summary from four countries. SOGial Biology, 39, 203-211. Smith, T. W. (1996). American sexual behavior: Trends, socio-demographic dif­ ferences, and risk behavior. In J. Garrison, M.D. Smith, & D. Bersharov (Eds.), The demography of sexual behavior (pp. 1-77). Menlo Park, CA: Kaiser Family Foundation. Smith. T. W. (1997). The impact of the presence of others on a respondent's answers to questions. International Journal of Public Opinion Research, 9, 33-47. Sonenstein, F. L., Pleck, J. H., & Ku, L. C. (1989). Sexual activity, condom use, and AIDS awareness arnong adolescent males. Family Pl4t�ning Perspectives, 21, 152-158. Sperber, D., & Wilson, D. (1986). Rekvance: Communication and cognition. Cambridge, MA: Harvard University Press. Srull, T. K., &: Wyer, R. S. (1979). The role of category accessibility in the interpreta tion of information about persons: Some deter111inants and implica­ tions. joumt�l of Personality and Social Psychology, 37, 1660-1672. Stalnaker, R. C. (1974). Pragmatic presuppositions. In M. K. Munitz & P. K. Unger (Eds.), Se�,antics and philosophy (pp. 197-214). ew York:New York University Press. Stapel, D. A., Martin, L. L., & Schwarz, N. (1998). The smell of bias: What instigates correction processes in social judgments? Personlllity and Social Psychology Bulletin, 24, 797-806. Stefanowska, M. ( 1977). The feeling of '*cultural inadequacyn and the validity of respondent's answers to questions about reading books. Studia Sociologi­ c:ne, 2, 133-143. Sternberg, S. (1969). Memory-scanning: Mental processes revealed by reaction titne experiments. Acta Psychologica, 60, 276-315. Stevens, S. S. (1975). Psytbophysics: Introduction to its peraptual, neural, and social prospects.New York:Wilcy. Stinson, L. L. (1997). Fi1141 r�ort: The subjeaive assessment of income and expmsu: Cognitive test results. Washington, DC: Bureau of Labor Statistics.

373

374

References Strack, F. (1992). Order effects in survey research: Activation and informative functions of preceding quations.ln N. Schwarz & S. Sudman (Eds.), Context effects in social and psychological (pp. 23-34). New York: Springer­ Verlag. Strack, F., & Marrin, L. (1987). judging, and co•nmunicating: A process account of context effects in attitude surveys. In H. Hippler, N. Schwarz, & S. Sudman (Eels.), Social infomration processing and survey tMth­ odology (pp. 123-148). New York: Springer-Verlag. Sttack., F., Martin, L., & Sch (1988). Priming and contartunication: The social detertninants of information usc in judgments of life satisfaction. EMro· ··

·

•

2. pean JourNal of Social Psychology, 18, 4 Strack, F., Schwarz, N., & Gschneidinger, E. (1985). Happiness and reminiscing: The role of time perspective, affec� and mode of thinking. ]out,,tal of Person­ ality and Social Psychology, 47, 1460-1469. Strack, F., Schwarz, N., & Wanke, M. (1991). Semantic and tic aspects of context effects in social and psychological research. Social Cognition, 9, 111-125. Stulginskas, j. V., Verreault, R., & Pless, I. B. (1985). A comparison of observed and reponed restraint use by children and adults. Accidmt Analysis & Preven­ tion, 17, 381-386. Suchman, L., & Jordan, B. (1990). Interactional troubles in face-to-face survey interviews. Journal of the American Statistical Association, 85, 232-241. Suchman, L., & jordan, B. (1992). Validity and the collaborative consauction of meaning in face·to·face surYeys. In J. M. Tanur (Ed.), Questions about questions: Inquiries into the cognitive bases of surveys (pp. 241-267). New ·

·

York: Russell Sage Foundation. Sudman, S., Bickart, B., Blair, J., & Menon, G. (1994). The effects of level of panicipa.rion on reports of behavior and attitudes by proxy reporters. In N. Schwartz & S. Sudman (Eds.), Autobiographie�JI naen•ory and validity of ret­ rospective reports (pp. 251-265). New York: Springer-Verlag. Sudman, S., & Bradburn, N. (1973). Effects of time and memory factors on response in surveys. journal of the American Statistical A�ocilltion, 68, 805-

815. Sudman, S., & Bradburn, N. ( 1974). Response effects in surveys: A review and synthesu. Chi�go: Aldine. Sudman, S., & Bradburtt, N. (1982). Aslt.mg qwstions: A practical guide to questionnaire design. San Francisco: Jossey-Bass. Sudman, S., Bradburn, N., 8c Sch N. (1996). Thinking about tmSUJ�: Th� application of cognitive processes to SJitvey ntethodology. San Francisco: jossey-Bas . proceSudman, S., Finn, A., & Lannom, L. (1984). The use of bounded dures in single interviews. Public Opmion Quarterly, 48, 520-524. Tanfer, K., & Cubbins, L.A. (1992). Coital frequency among single women: Normative constraints and situational opportunities Journal of Sex Research, ..

29, 221-250. Tangney, J.P., ac Fischer, K. (1995). Self-conscious emotions: Shame, pilt, embarrassment, and pride. New York: Guilford Press.

I

r

I I

eferences Tangney, J. P.,

i1ler R.

·

·

..

, Flicker, L .. ,

Barlow D. H (1996). Are hame, ..

guilt and embarra ment di tin t emotion ? journal of Personality and . ocial Psychology, 70 1256-12.69. Tannenhau

. K., Boland, J. E.,

auner, G. A.,

Carl on, G.

(1 93). More on combinatory lexical informati.on: Tbemati tructure in par ing and interpr tation. In G. ltmann R. Shillcock (Eds.) Cognitive ttJodels of speech processing (pp. 297-319). Hill dale, I j: Erlbaum. (1978). Self-generat d attitude chang . In . B rkowitz ( d.) Ad­ Te er, vances in experimental social psychology ( · ol. 1, pp. 289-338). e York: Academic Press. Tess r, A., & L on , C .. (1977} .. Cognitiv. schema and thought a d termin­ aot of attitude chang . Journal of xperin1ental ocial Psychology 13. 340356. Te ser, A., Rosen, . (1975). The reluctance to tran mit bad news. In L. Berkowitz (Ed.) Advances i11 experitnental social psychology (Vol. 8 pp·. 193232). ew York: Academi Pre . Brow , I. Jr·. (1974). Time pe ception and th filledThoma , E.... C., duration iillu ion. Perception & Psyc.hophysics, 16, 449-458. C. P. (1982). · emory for unique personal event : The roommat .,...hompson, . study. Menso·ry and Cognition·, 1 0 J.24-332. Betz, A. L. (1996). AutobioThompson, · . P., Skowronski,J.J., Lar en, . . g,;aph.ical 111ert1ory. Mahwah, J: rlbaum. as ey, J (1988). Trends in 'United tate telephone co¥ rThornberry, 0. age across time and ubgroups. In R. Grove P. Biemer L. Lyberg j. . a sey, W. ichoUs & J. Waksberg (Eds.) Telephone survey metl1odology (pp. 2549). ew York: John· iley. Thur ton , L. (19·27). A law of comparativ judgm nt. Psychological Review 34 273-286. Touran au R. ( 984). Cognitive cience and urvey method . InT. Jabine, M. Straf, J. Tanur R. Tourangeau (Eds.) Cogn'itive aspects of 5urvey design: ashington, DC: - a­ Building a bridge between disciplines (pp 73-100). tiona! Academy Pre . Tourangeau R. (1987). ttirude measurement: A cognitive p-rspecriv . In H .. Hippler, . hwarz & .. udman {Eds.), Social information processing and w York: Springer-Verlag. survey methodology (pp. 149-162).. Tourangeau R. (1990). Comment. journal of the American tatistical Associa­ tion, 85, 250-251. Tourangeau- R (1992). Context effects on attitude respon e : The role of re­ trieval a d memory structure . In r Schwarz & S. udman (Ed .), Context effects in social and psychological research (pp. 35--47).. ew York: Springer­ Verlag. Tourangeau, R. asin ki, (1986). Context effects in attitu·de surveys. Unpublished manu cript. Tourangeau, R., & Ra in ki, K. (19.88). Cognitiv_ proce e underlying co text effects, in attitude measurement. Psychological Bulletin, 103 299314. Tourang au R .. , Ra inski, K., & Bradburn, . (1991). asuring happine in ·.

·

·

·

..

·

..

.

..

•

·

·

·.

375

376

References surveys: A test of the subtraction hypothesis. Public Opinion Quarterly, 55, 255-266. Tourangeau, R., Rasinski, K., Bradburn, N., 8c D'Andrade,R. (1989a). Carry­ over effects in attitude surveys. Public Opinion Q��arterly, 53,495-524. Tourangeau, R., Rasinski, K., Bradburn, N., & D'Andrade, R. (1989b). Belief accessibility and context effects in attitude measurement. Journal of &pe•i­ mental Soci41 Psychology, 25, 401-421. Tourangeau, R.. , Rasinski,K., & D'Andrade, R. (1991). Attitude struct\lre and belief accessibility. journal of Experimental Social Psychology, 27, 48-75. Tourangeau, R., Rasinski, K., Jobe, j. B., Smith, T. W., & Pratt, W. (1997). Sources of error in a survey of sexual behavior. journal of OfPcial SttJtistics, 13._ 341-365. Tourangeau, R.,Shapiro,G.,Kearney, A.,& Ernst, L. (1997). Who lives here? Survey undercoverage and household roster questions. ]ounu�l of OfPcUII Stlltistics, 13, 1-18. Tourangeau, R., & Smith, T. W. (1996). Asking sensitive questions: The impact context. Public Opin­ of data collection mode,question fortnat, and q ion Quarterly, 60, 275-304. Tourangeau, R., & Smith, T. W. (1998). Collecting sensitive informa tion with different modes of data collection. InM.P. Couper, R.P. Baker, j. Bethlehem, C. Z. Clark, J.M.anin, W. L. Nicholls, 8c J. O'Reilly (Eds.), assisted survey information collectiort (pp. 431-454). ew York: Wiley. Tourangeau, R., Smith, T. W., & Rasinski9 K. A. (1997). Motivation to repon sensitive behaviors on surveys: Evidence from a pipeline experiment. Journal of Applied Social Psychology, 27,209-222. Traugott, M W., & Katosh, J.P. (1979). Response validity in surveys of voting behavior. Public Opinion Quartnly, 43, 35'-377. in dispositional attriTrope, Y. (1986). Identification and inferential pr bution. Psychological Review, 93, 239-257. Tulving, E. (1983). Elmrents of episodic "aemory. Oxford: Oxford Univenity Press. Tulving, E. (1984). Relations among components and processes of memory. BehaviortJI 11nd Broin Sc""as, 7, 257-263. Tulving, E. (1985). Memory and consciousness. Canadian Psychology� 26, 112. Tulving, E., & Thon1son, D.M. (1973). Encoding specificity and retrieval pro­ cesses in episodic memory. Psychological Rwiew, 80, 352-373. Turner,C. F., Ku, L., Rogers, S.M., Lindberg, L. D., Pleck,j. H., & Sonenstein, F. l. (1998). Adolescen t sexual behavior, drug use, and violence: reporting with computer survey technology. Scimu, 280, 867-873. Turner. C. F., I ,j. T., & Devore,j. (1992). Effects of mode of administtation and wording on reporting of drug use. In C. Turner, j. l.essler, &: J. Gfroerer (Eds.), Survey measu re•r.mt of drMg �: MnhodologiCIIl studies (pp. 177-220). Rockville, MD: National Institute on Drug Abuse. Turner, C. F., Lessler, J. T.,&: Gfroerer,J. (1992). Surwy �asurernent of drug use: Methodological studies. RockviUe, MD : National Institute on Drug Abuse. ·

·

·

·

..

I

I I I

R Turner, C. F., &: Martin, E. (1984). Surveying subjeaive pheno�Mna. New York: Russell Sage Foundation. Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and probability. Cognitive Psychology, S, 207-232. Tversky, A., 8c Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185, 1124-1131. Tversky, A., & Kahneman, D. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47, 263-291. Tversky, A., & Kahneman, D. (1982}. Judgments of and by representativeness. In D. P. Slavic, & A. Tversky (Eds), Judgment under unartainty: Heuristics and biases (pp. 84-98). Catnbridge: Cambridge University Press. Tversky, A., &: Koehler, D. J. (1994). Support theory: A nonextensional repre­ sentation of subjective probability. Psychological Review, 101, 547-567. Tversky, B., & Tuchin, M. (1989). A conciliation of the evidence in eyewitness testimony: Conunents on McCloskey and Zaragoza. Journal of Experimmttzl Psychology: General, 118, 86-91. Udry, J. R. (1980). Changes in the frequency of marital intercourse from panel data. Archives of Sexual Behtltlior, 9, 319-325. Underwood, B. j., Zim111enltan, J., & Freund, J. S. (1971). Retention of fre­ quency inforanation with observations on recognition and recall. journal of Experi,Nntal Psychology, 87, 149-162. Underwood, G. (1975). Attention and the perception of duration during encod­ ing and retrieval. Peraption, 4, 291-296. Upchurch, D. M., Weisman, C. S., Shepherd, M., Brookmeyer, R., Fox, R., Celentano, D. D., Colletta, L.• & Hook, E. W., m. (1991). lnterpart11er relia­ bility of reporting of recent sexual behaviors. AmeriCiln ]cn�rlldl of Epidemi­ ology, 134, 1159-1166. Usher, J. A., & Neisser, U. (1993). Childhood amnesia and the · ·. · gs of memory for four early life events. Journal of Experi1nental Psychology: Gen­ eral, 122, 155-165. Von Neurnann, j., & Morgenstern, 0. (1947). Theory of games and economic behavior (2nd ed.). Princeton, NJ: Princeton University Press. Wadsworth, J., Johnson, A. M., Wellings, K., & Field, J. (1996). What's in a mean? An exan1inarion of the inconsistency between men and women in reporting sexual ·ps. Journal of the Royal Statistical Society, 159, 111-123. Wagenaar, W. A. (1986). My memory: A study of autobiographical memory over six yean. Cognitive Psychology, 18, 225-252. Walker, j. H., Sproull, L., &: Subramani, M. (1994). Using a human face in an interface. Proceedings of the Conference on Human Faaors in Comfnlters '94, (pp. 85-91). Boston: ACM. Wallsten, T. S., Budescu, D. V., & Zwick, R. (1992). Com� the calibration and coherence of ntunerical and verbal probability judgments. Managmunt Science, 39, 176-190. Warner, S. (1965). Randomized response: A survey technique for eliminating evasive answer bias� Journal of the Am�'ican Statistical Association, 60, 6369. ·

·

377

Warriner, G. K., McDougall, G. H. G., & Claxton, J.D. (1984). Any data or none at all? Living with inaccuracies in self-reports of residential energy con­ sumption. EnviroMUnt and Behavior, 16, 502-526. Waterton, J., & Duffy, J. (1984). A comparison of computer techniques and traditional methods for the coUcction of self-report alcohol con­ ·

·

·

surnption data in a field sW'Vey. InterJUJtional Statistiul Rer.Mw, 52, 173-182. Waterwonh, J. A. (1985). Memory··.·· and rhe psychophysical scaling ·

of duration. Perception, 14, 81-92. Watkins, M. j., & Kerkar, S. P. (1985). of a twice-prcsellted ite1n without recall of either presentation. journal of Menrory and Language) 24, 666-678. as a basis for Watkins, M. J., & LeCompte, D. C. (1991). Inadequacy of frequency knowledge. journal of &perinrentiJI Psychology: 1-eat"Hfng, Mem­ ory, & Cognition, 17, 1161-1176. Wedell, D. H., Parducci, A., & Geiselman, R. E. (1987). A analysis of ratings of physical a : Successive contrast and simultaneous assimilation. Journal of Experil1rental Socilll Psychology, 23, 230-249. Weeks, M. ( 1992). Computer-assisted survey information collection: A review of CASIC methods and their implications for survey operations. journal of O{Pcial Statistics, 9, 445-465. Wegener, D. T., & Petty, R. E. (1995). Flexible correction processes in social judgment: The role of naive theories in corrections for perceived bias. Journal of Pet�onality and Social Psychology, 68, 36-51. Wells, G., & Gavanski, I. (1989). Mental simulation of causality. }oNnaal of Personality and Social Psychology, 56, 161-169. ·

Whisman, M.A., & Allan, L. E. (1996). Attachment and social copition theo­ ries of romantic relationships: Convergent or complementary perspectives. journal of Social and Personal Relationships=- 13, 263-278. White, R. T. (1982). Memory for personal events. Human Learning, 1, 171183. Whitten, W. B., & Leonard, J. M. (1981). D irected search through autobio­ graphical memory. Mt,tory & Cognition, 9, 566-519. Whittlesea, 8 W.A. (1993). Dlusions of familiarity. Journal of Experimental Psychology: Learning, M�'nory, and Cognition, 19, 1235-1253. Wickelgren, W. A. (1973). The long and short of memory. Psychological BNII�­ tin, 80, 425-438. Wickelgren, W. A. (1974). Single-trace fragility theory of memory dynamics. Memory & Cognition, 2, 775-780. Williams, M. D., & Hollan, j. D. (1981). The process of retrieval from very long-term memory. Cognitiv� Scimce, S, 87-119. Willis, G. (1997a). The use of the psychological laboratory to study sensitive topics. In L. Harrison &: A. Hughes (Eds.), The wlidit)' of self-reported dTNg IUe: Improving the ace14racy of SIWVry �stimates (pp. 416-4.38). NIDA Mono­ graph 167. Rockville, MD: National Institute on DrugAbuse. Willis, G. (1997b). NCHS Cognitive Interviewing Proj�ct: Ger.ral coding sc.henu for qu�stionnaire problems. Hyattsville, MD: National Center for Health Statistics. Willis, G.7 Brittingham, A., Lee, L., Tourangeau, R., & Ching, P. (1999). Re· spon.st! �rrors in s11n�eys of children�s immumutions.Vital and Health Statis-

I 1

•

I

References

I tics, Series 6,

3 79

wober 8. Hyattsville, MD: National Center for Health Statis·

•

acs.

Willi� G., DeMaio, T., & Harris-Kojetin� B. (1999). Is the bandwagon headed to the methodological promised land? Evaluating the validity of cognitive interviewing techniques. In M. Sirken, D. J. Herrntann, S. Schechter, .. Schwarz, j. Tanur, & R. Tourangeau (Eds.), Cognition and survey research (pp. 133-154). New York: Wiley. illis, G., asinski, K., & Baldwin, A. (1998). Cognitiv� research on responses to sensitive survey questions (Working Paper Series, o. 24). Hyattsville, MD : arional Center for Health Statistics, Cognitive Methods Staff. Willis, G., Royston, P., & Bcrcini, D. (1991). The use· of verbal report methods in the development and testing of survey questionnaires. Applied Cognitive Psychology, 5, 251-267·. WilJjs, G. B., & Schechter, S. (1997). Evaluation of cognitive interviewing tech­

niques: Do the result generalize to the field? Bulletin tk Methodologie Sociol­ ogique, 55, 40 66. Willis, G., Sirken., M., & athan, G. (1994). The cognitive aspects of responses to sensitive survey questions (Working Paper Serie , No. 9). Hyattsville, MD : arional Center for Health Statistics, Cognitive Method Staff. Wilson, T. D., & Brekke (1994) Mental contamination and mental correction: Unwanted influences on judgments and evaluation. PsychologiC�Jl Bulletin, 116,117-142. Wilson, T. D., & Dunn, D. (1986). Effects of introspection on attitude-behavior consistency: Analyzing reasons versus focusing on feelings. journal of Experi­ m•ntal Social Psychology, 22, 249-263. Wilson, T. D., & Hodges, S (1992). Attitudes as temporary con tructions ln L. Martin & A. Tesser (Eds.), The construction of social ;udgments (pp. 37-66). ew York: Springer-Verlag. Wilson, T. D., Hodges, S., & LaFleur, S. (1995). Effect of introspecting about reasons: Inferring attitud s from acces ible thoughts. joutnal of Personality and Social Psychology, 69, 16-28. (1996). A new look Wilson, T. D., Houston, C. E., Etling, K. M., & Brekke, at anchoring effects: Basic anchoring and its antecedents. Jounu:�l of Exptri· mental Psychology: Gf!neral, 1 25, 387-402. Wilson, T. D., Kraft, D., & Dunn, D. (1989) The disruptive effects of e plaining attitudes: The moderating effect of knowledge about the attitude object. Jour­ nal of Experimental Social Psychology, 25, 379-400. Wit n, T. D., LaFle ur, S. J., 8c Ander on, D. A. (1995). Th v lidity nd con­ sequences of verbal reports about attitudes. In . hwarz & S. Sudman (Eds.), Answering questions: Methodology for deter1P1ining cognitive and com­ municative processes in survey research (pp. 91-114). San fr ncisco: Jossey­ Bass. Woodberry, R (1997). The missing Pfty percent: Accounting for the gap be­ tween survey estimates and h�ad counts of church attendance. Master's th i , Sociology Deparanent, un·versity of otre Dame, South Bend, · . Woodrow, H. (1951). Time perception. In S. S. tevens (Ed.), Handbook of experimental psychology (pp. 1224-1236). ew York: Wil y. Wyer, R., & Hartwick, J. (1984). The recall and use of belief tatements as bases ..

..

..

.

..

..

I

380

References for judgments: Some detertninants and implications. journal of &perimmtal Soeial P$Ychology, 20, 65-85. . (t9n). Some funber evidence for the Socratic effect Wyer, R., & Rosen, using a subjective probability model of cognitive · rion. ]OMntal of Personality and Social Psychology, 24, 420-424. Wyner, G. A. (1980). Response errors in self-reported nurnber of arrests. Socio­ logical Methods and Rnearch11 9, 161-177. Ya1runarino, F .. J., Skinner, S. J., & Childers, T. L. (1991). Understanding mail survey response behavior. Public Opinion QuartBrly, SS, 613-639. Young, N. (1989). Wave-seam effects in the S IPP. Prouedmgs of the Seaion on Survey Research Methods, A•rur1can Statistiad (pp. 393-398). ahon. Alexandria, VA: American Statistical Yzerbyt, V. Y., Schadron, G., Leyens, J.·P., & Rocher, S. (1994). Social judge· ability: The impact of meta·infor·111ational cues on the use of stereotypes. Journal of Personality and Social Psychology, 66, 48-SS. ·tive theories in social psychology. In G. Lindzey &: Zajonc, R. B. (1968). E. Aronson {Eds.), Handbooll of social psychology (2nd ed., Vol. 1, pp. 32� 411). Readin� MA! Addison-Wesley. Zakay, D. (1990). The evasive an of subjective time meaaureJJJent: Some meth­ odological clile1nrnas. In R. A. Block (Ed.), Cognitiv� models of psychologiad time (pp. 59-84 ). Hillsdale, NJ: Erlbaunt. Zakay, D., Tsal, Y., Moses, M., & Shahar, I. (1994). The role of .tation · in p�ospective and retrospective titne Memory & Cognition. 22, 3 .351. Zaller, j. R. (1992). The natllre and origins of mass opinion.. bridge University Press. Zaller, j., R., ac Feldman, S. (1992). A simple theory of the survey response: Answering questions venus revealing preferences. American joJirnal of Politi­ ca1Science.36,S79-616. Zaragoza, M.S., & McCloskey, M. (1989). Misleading postevent infortttatioo and the memory impairment hypothesis: Cotnment on Belli and reply to Tver· ology: General, 118, 92-99. sky and Tuchin. journal of E:xperime�•tal P ·

.

I

•

Author Index I

Abbott, V., 6.2 Abelson, R.,69, 78, 79, 81, 172,

Bargh, J., 169 Barker,K., ill Barlow, D., 285, 306

202,173,274

Abernathy, J.,272 Abramson, P., 273, 274,280, 187 . .

Barsalou, L., 69, 70. 78. 79, 96,

lli

H.1 271

Allan, L., 1 17, lli Allen,J.,3, 36, llO Allport, G., 167 Alwin, D., 17, 249, 250, 251,2521 293,304,1ll Anderson,B.,273, 274,280, 287 Anderson,D., 179,2�08

·

Armstrong,J.,95, 162, 16.3 Armstrong, S., � Ayidiya,S., 2100

..

Ball-Rokeach,S., lli

Ba ,R.,261,307 Beatty, P., 6.6 ·

, 1., lli Bekerian, D , 80 Bell, C., 83, M Bell,K., 285, 286, 287 Belli,R.,42, 274 Belson, W., 24, 25 Bercini,D., 19, 323 Berent,M., ill Bergman,L., 303

Bachrach,C., 276 Baddeley,A., 77, 84, 115, 132, 131 Bahrick, H., 83, � Bahrick,P., 83, M Bailar,B., 319 Balliet,S., 66 Baker, R.,299! 300,3011 303,,, 308 Ba l, D., 169 Baldwin,A., 233, 235, 265, 271, 281, 282, 283, 284,2.21

Bauman, K., 271

..

Anderson,J.,14,32, 772 87, 912 207 Anderson,N., 11, 13, 181, 210 Anderson,R.,176 --.r D., 270,1H Aquilino, W., 270,281, 294,296 •

Bartlett,F., 22 Barzelay, D., 270 Bassili,J.,14, 217, 325-326, 331

, c., 269 Bershad,M., 319,320, 321, YO Bettman, J.,8 Betz, A., 11, 67, !Jl, 92, 112, 114, 115, l16, 1211133,244 Bickart,B.,22, 65, 66, 117, 147, 151, 328. 329 Biedertnan,A., ill Riemer, P., 300, 302,308, 314, 3170 Bienas, j.,248 Biernat,M., 212, 221, 2�22, 228 Binnick ,R., 101 Binson,D., lli

381

Copyrighted rna ri

I

382

Author Index Bishop,G., 170,200,207,216,246, 310,311 Bjork,R., 63 Black,J.,69, 78, 79,81, 326

Brown, 1., Jr.,118

Blair,E., 12,21,86,94,139,146, 148,149,151,152,154,155, 156,159,160,163,337 Blair,J.,22, 65. 66, 147, 151, 307, 328,329,330,331,332,333 Bless,H., 150,156,160,201,207, 208,209,211,212,216,221, 224,225,226,337 Block,R., 118 Blower,S., 276 Bohner,G., 150 Boland,J., 37 Bolinger,D.,26,30 Bolton, R., 328,329 Boltz, M., 118 Botrunareddy,S., 32

Brown, P., 287 Bruce, D., 142, 145 Bruner,J.,71 Brunner,G., 307 Budescu,D., 47,162 Burrelli,J.,291 Burt,C.,112-113,114,115,119, 120,133 Burton,S., 12,21,86, 94, 139,146, 149,151,152,154,155,156, 157,160,163,337

Bonham, G., 272 Boone, S., 118 Borg, 1., 119 Bower,G., 69 Bowker, D.,210 Bradburn,N., 2,22,42,47,49,60, 66, 81,83, 86, 87,88, 92,94, 120,121, 129-130,132,133,137, 145,146,147,174,179,180, 182,197,199, 201,203,204, 205,206,207,209,224,225, 228, 234,235, 236,237,238, 251,257,259, 260,268,269, 280,295,299, 300,301,303, 308,314,315,335,337 Branden,L., 270,280,296 Bransford,J.,202 Breckler,S., 185 Brekke,N., 144, 207 Brewer, M.,324 Brewer,W., 68,84,85, 96 Brittingham,A.,66y 146,149,152, 269,271,339 Brook,L., 203 Brookmeyer, R., 276 Brooks, C., 319 Brown, D., 216, 337

Brown,N., 10,79,88,137,146, 147,149,150,151,152,154, 155,156,151,278,337

Caces,M.,270,280,296 Cacioppo,J.,7,167,252,307 Calder, B.,141 Camburn, D.,335 Campanelli, P., 351 Canchola,J.,276 Cannell,C.,5,6,7,16,17,85,92, 94, 301,311,314 Card,J.,269 Carlson,G., 37 Carlson,J.,176,180,203,209 Carpa1ter,P., 25,36,40 Carroll,S., 307 Caner,W., 262 Cashman,E., 137,146,147,150, 151,152,154,155,156,157,337 Catania,J.,261,276 Catlin, 0., 299,300,301 Cavi� S., 269 Celentano,D.,276 Centers,R., 240,241 Chaiken,S., 7,169,171,184,302 Chang, P., 212 Chaves,M.,274 Cheadle,A.,269,271 Chen,D.,168 Childers,T., 261,307 Ching,P., 146,149,152,339 Chitwood,D.,261 Chou,C.,270 Chu,A.,86,97,155

Author Index Cialdini, R., ll8 Claggett, W., 274 ...... Clark, A., 269, 277 Clark, F., 241-243 Clark, H., 9, 53, 54, 55, 56, 57, 202 Claxton, J., 169 Clayton, R., 290, 294 Clore, G., 177, 200, 2�07, 208 Coates, 0., 261 Coates, T., 261, 267 Cobb, w., 286 .. an, w., 2 Coder, J., 272 Cohen, N., 8l Colletta, L., 176 Collins, A., 10, 43, 77, 110, 157, 159, 2o3\ ru Combs, B., 161, 162 Conrad, F., 19, 55, 56, 137, 146, 147, 150, 151, 152, 154, 155, 156, 157, 298, 319, 326, 327, 328, 329, 333, 337 Conradt, J., 310 Converse, J., 39, 40, 43, 60, 28.6 Converse, P., 13, 1,2, 169 Conway, M., 70, 15-76, 79, 80, 110, �

ill

Couper, M., 259, 261, 262, 263, 279, 289, 290, 291, 301, 309, 338 Craik, F., 142, 145 Crain, S., 1Z. Crelia, R., 180, 220, 221 Crowder, R., l18 Crowne, D., 258 Cubbins, L., 2i6 Curran, T., 138, 1391 142, 157 Curtice, J., 299, 300, 301 Daamen, D., 2.40 D'Andrade, R., 14, 176, 177, 179, 182, 199, 206, 207, 325 D ·.. C., 174 Davis, W., 161 deBie, S., 240 DeMaio, T., 232 258, llZ enu· W., 2 Denniston, W., Jr., 95, 1622 163

38 3

Dent, C., 271 DePaulo, B., 285, 286, 287 Deutsch, B., 150, 247, 248 Devine, P., 269 Devore, j., 270, 280, 226 Diehr, P., 269, 271 Dillman, D., 26, 309, 110 Dominitz, J., 160, 161, 162, 16.3 Donaldson, M., 12, 21, 78, 85, 95, 96, 13:4. Dovidio, J., 2,69 Dowd, K., 262 Downing, J., 167, 176, 177, 206,

325 Dowty, D., 101 Drake, R., 167, 176, 177, 206, 3li Duffer, A., Jr., 271, 2&1 Duffy, j., 269, l01 Dull, V., 324 Duncan, G., 82, 85, 8.6 Dunn, D., 171, 179, W Eagly, A., 30l Eisenhower, D., 86, 97, 155 Ellison, S., .276 Ellsworth, P., 128 Epstein, J., 285, 187 Ericsson, K., 3-26, � Ernst, L., 1422.8l Esposito, J., 19, 94� 95, 281, 324, 336tllZ Erling, K.,­ Evans, N., 269 Evans, V., 21:76 Everitt, B., 324 Fabrigar, L., '-49 Farr, J., 240, 241 Fathi, D., 93, 336, 337 Fay, R., 262 Fazio, R., 12, 167, 168, 169, 1 n, 173, 177, 178, 179, 203, 221, 222!247!2692325,337 Fecso, R., 2.91 Feigenbaunt, E., Zl Fein, D., 281 Fetcher, E., H1

Copyrighted rna ri

I

3 84

Author Index

Feldman, J., 159, 1751 185,186, 187, ·ss, 189, 190,286 Fcndrich, M., 269, 270 Ferber, R., 307 Ferguson, R., 120, 121 Fiedler, j.,82, 85, ,31 336, llZ

field, J., m Fillmore, C., :2.3, 15, � Finn, A., 82 Fischer, G., 16.3 Fischer, K., ill Fischhoff, B., 12, 18,161,162, l1S Fletcher, J., H Flicker, L., 285, 306 Fodor, J., 32, 1Z Fogg, B., 302 Forrest, J., 265, 268,� Fonyth, B., �9, 319,328, 330 Fowler, F., 38� 40,55, 311, 325, 331,-......-.

Fox, R., 226 Freedman, D., ill

Freund, J., 138! 139 Friedman, W., 11, 112, 1 4, 335 Fuhrman, R., 80.

Fujii, E., 273 Gagnon, J., 256 , 180 Galambos, J., 62 Galinat, W., ill Galitz, W., 309 Ganesh, G., 148 Gaskell, G., 150, lli Gavanski, I., 212 Geiselman, R., 212, 220,223 Gentner, D.y 10, 43, 157 Gerber, E., 324 Gfroerer, J., 256, 2,96 Gibson, D., 261 Gigerenzer, G., �45, 16Q Gillund, G., 82 Gilpin, E., 269 Gleitman, H., � Gleitman, L., 46. Golding, j., 11 Goldsmith, L., 67� 70, �st ill Goldstein, D., 145

Goldstein, H., ill Gordon, M., 95, 162, W Govender, R., .16.2 Graesser, A., 3, � �9, 32, 33 , 412 ll Greeat, N., l02 Greenberg,B., lli Greenwald, A., 273, 274 Grice, P., 9, so-s3, 54,202, 203, .

.

204,205,209

Groenendijk, J., 2Z Gropper, B., 270 Gross, S., 118 Gfioves, R., 4, 261,268,294,�95, 300,301. 314,319,322,323, ,338 Gschneidinger, E., 176,207,21 ·, 212, 2],,3, 226

Hackett-Renner, C., 41 Hadaway, K., 274 Hainer, P., 256, 28.1 -·

M., 2,4, 319,3202 321,

Hanson, C., 61 Harrell, L., 290,22.1 Harris, G., ill Harris-Kojetin, B., 23, 327 Harrison, L., 270 Harsch, N., 81 Hart, c., 286 Hartwick, J., rn Harvey, 0., 331 Hasher, l., . 39, 140, 1A1 Hashtroudi, S., Z& Hastie, R., 7,13, 173, 179, 314

Hatche� S., 286 Hauck, W., 2776 Hay, M., 86,87, l..SS Heberlein, T., 261,307 Hedges, L., .7.2, I 16,120,121,132, 235

llZ Hennessy, M., 2.73 Henrion, M., 163. Herold, E., 276 Herr, P., 203, 221, 222, 337 Herrn1ann, D., �9, .21, 66, 7l, .82t '

238

'

'

'

.

'

.

298,319

Higginbotham, J., 21,29, ll

Copy righted m a rial

Author Index E., 207, llZ Hild, T., 150,156, 160 Hill, D., l2l Hilton, D., 21z Hines, C., 256! 28.1 Hin . n, D., 72,138,139,141, 142,144, llZ Hippler, H., 150,200,2321 241243,247,248,252,253,262, 288,293,310,3 1,314

Hirst, w. , 6.8

Hocbstim, J., 294 Hockley, W., 142, � Hodges, S., 167, 1LZ , 194,197, 208, 334 Hofferth, S., 2Z6 Hollan, J., � Holmes, D., 9.2

Holyoak, K., 246 Honaker, L., 303 Hook, E., 2.76 Hornik, j., U1 Horvitz, D., 272: Hougland, J., 294 Houston, C., Hovland, C., 337 Howell, W. , 140 Hser, Y., 270 Hubbard, M., 19, 300,302, 308 Huber, J., 29,iS HuH, j., 217 Hughes, A., 296 Hurwitz, W., b 319,320,321, YO Huttenlocher, j., '?2, 1161 120,121, 132,234,235,238,337 Hyman, H., 210,286,337

Hymes, C., 16.2 Ingram, S., 299,300,l0.1

Jabine, T., 122, 313,319,335 Jackendoff, R., 32 Jacoby. L.. HJ jenkins, C., 26, 309 J L., 303 jobe, j., � 19, 70, 85,86,93, 94, .

95,97,126,148,233,234,235,

38 S

263,265,271, 281,282, 283, •

-

•

0

284,291,297,298,299,300, 301,305 ,323,324,326,336, 337 ..

johnson, A., 2.Z6. johnson, B., 3,2 Johnson, E., t 2ll johnson, E. C., � johnson, L., 269,270 Johnson, M., 78, 82,202,262. johnson, N., 202 Johnson, R., 299, 300,301,303,l08 johnson, T., 294 jones, C., 207, 3..31 jones, E. E., 171,�69 Jones, E. F., 265,268,269 jonides, J., 139, 141, H2 Jordan, B., SS, 57,311,319,338 Judd, c., 167J 169,176, 1 n, 206, 217, 325

Jungeblut, A., 3Jll just, M., 36,� Juster, T., 16_0 Kahn, J. , 21.6 Kahn, R., 294 Kabneman, D.,

10,125,137,139, 143,144, 145 ,151,161, 16.2 Kalsbeek, W., 319 Kaltoo, G., 124,125, . 26,203

Kane, E., 286,2;87 Kang, S 270 Kaplan, K., 18.8 .•

Kardes, F., 167,168,179, 12.5 Kashy, D., 285,287 Kasprzyk, D., W

Katosh, J., 274 Katz, D., 2�86 Kautz, H., 110 . Kay, w., 66, 269,271 Kearney, A., 41 2.&1 Keenan, j., 66 KeUey, C., 93,3361 .337 Kennickell, A., 234,264 Kenny, D., W Kerkar, S., W Kerwin, J., 66 Kiesler, S., 290,299,JOt

Copyrighted rna ri

I

3 86

Author IDdex Kinder,D., 175 King, K., 122 Kinne, S., 269,271 Kirkenol, S., 285, 287 Kirsh,I., 303 Klassen,A., 216. Kleinman,S., ZO Kline,D., 118 Klinger,M.,�2,�5, 93,336,.337 Klurnpp,G.,208 K.nauper,B., 241�243,212 Knibbe,R., 269 Knowles,E., 228,.222 Koehler,D., 119l 160, 162, W Koepsell,T., �69,271 Kolodner,J.,70,73-75,21 Kolstad, A., 10l Kraft,D., l71t �79, 18.5. Krensky,L.,Ztl Kristiansson,K., lOJ Krosnick,j.,1 ! 167! 169117111721 176,177, �06,�17,�49,zso. 0

251,Z52�ZS3�293,3Q4,32S Krueger,R., 324 Ku, L., 2 5 6,276,290, 300, 308 Kubovy,M.,ill Kuklinski, J., 179, 201, 222 Kulka,R., 259� 262,263 Kurbat, M.,89,90,91,114,116, 2M 0

'

0

.

'

LaFleur,S., 177, 179, 208, 3M Lakoff,G.,234 Landauer,T., 6J Landy, F., 240, 241 Lanier,A., ill Lannom,L., 82 l.arsen,S., 11, 22,65:�. �1192,1121 114,115,116, 133, 244, Lanon,R., ll Lasbeek,W., 176 Lau, R., 240, 241 Lau1uann, E., 256,2_8__0 Laurent,A., .S Lavine,H., 2,17 Layman,M.,1611 162 LeCompte,D., 142,145! W

Lee, L.,146,149,152, 339 Lehnert, w., � 2

I.emmens, P., 262 Leonard, J., 83,84,� I,eone, C., 171, 184 I.essler, j., 12, 19, 43, 147, 148,149, 151,256,270, 271, 280,281, 296,300,302,308, 319,326, �28,330, 331, ;336,�37, Levinson,S., 2�87 Levitt, E., 276 Lewis,D., 9,42. 10.8 Lewis, V., 84,115,ill Leyens,J.,1l Lichteustein, S., ?S, �61, �62, 163. Lindberg, L., 276,290, 300 , 308 Linde, C., Z1 Lindsay, D., Z8 Linton, M., 78,13 Linville,P., 111 Locander, W., 268, ill lEe. M., � 73, 1M Loftus, E.., 12,lL 42,78,821 85, 89,91, 93,95,96.118.134,146, 148,154,155,160,163,273, 274, 326,335,336,l1Z London, K.,271 Lopes, L., 18tl LoScuito,L.,270,lli Ludwig, J.,2102 21l lui, L., 324 Luker, K., 1H Lyberg,L., 314 Lynch, j., m '

Macaulay, L.,286,281 MacGregor,D., �S, 163.

Madigan, s., � Madow,W., 2 Madson, C., 25 Magura, S., 270 Mah, w., 246 Mai, H., 180,203, 204, 205, 209, 223 Mak,j., ill Malt, B., 46 0

-

. 'J., 202

Copyrighted rna ri

I

Author Index Mangione, T ., iS Manis, M., 212,1.21, 222, 228 . c.• 291 Manski, C., 160, 161, 162, l6J Marburger, W., 89,336, 337 Marler, P., 274 Marlowe, D., 258 Marquis, K , � 125, 126, ill Martin, E., 203, 256, 281, 3 1 3, 324, 325 Martin, J., 299! 300,301, 30.3 MaJ'tin, L., L 17, 18,20,172, 178, 180,201,203,207, 209,213, 220,221, 222, 314, 319,337 Martin, P., 110, l21 ..

... . i, A., 12J Mason, R., 1762 180, 2031 202 'j.,293

Mathiowetz, N., 82, 85, 86, 259, 262,263,279,300, 301, 314 Mauner, G., 31 I, D., ill May, R., 27�6 McCauley, R., Zl McClelland, J., 16 McClendon, M., 176, 200,204 McCloskey, M., 42, 81 McDonel, E., 1.6.8 McDougall, G., 2.69 McGill, A., 212 McGonagle., K., 274 McGraw, K., 173, 174 McGuire, W., 2131 287 McMahen, C., � i McMullen, M., 2111 221,2.22 McNaughton, B., 1Ji McQueen, D., 294 B., 11,21,78,85,91,94, 95,96, 134,146, 148,154J 155, t60, 163, 335, ll6 Menon, G., 22, 65,66, 146, 147, 148,: 150! 151,155,156,159! 160, 163,3281 329,337 Michael, R., 256, 18{} Michaels, S., 256, 2BO Michalski, R., 110, m Mieczkowski, T., 270

387

Milburn, J., 169. Miles, C., 47, 101 Millar, M., 185 Miller, E., 262,28.8 Miller, M., 124, 1.25, 126 Miller, P., S, 6, 16,85, 92,94,268, 301, 314 Miller, R., 185,306 Mingay, D., 66, 70, 93, 148, 228, 251,322,326,336,337 Minerer, J., 112 Moon, Y., 302 Moore, J., 67, 123, 125, 126, 264, 268,269,272, ill Morganstein, D., 86t 87t ill Morgcnstcr11, 0., 28.1 Morris, M., 69, 2Z8 Moses, M., 118 Mosher, W., 271,181 Moss, L., ill Mott, F., 271 Moxey, L., 47, 42 Mueller, J., 174 Mullen, P., 298,319 Mullin, T., 19, 16.3 Murphy, G.,� Murray, D., 265, 268,269,271 Myaskovsky, L., ill Myers, S., 270 Narayan, S., 252,153 Nass, C., 302 Nathan, G., 281,282 Nebes, R., 81 Neisser, U., 67, 71,8.1 Nelson, T., 212, 221,].22,228 Neter, j., 11, 86, 881 891 92, 97, 126, 127, 128,129,130, 133, 146t 155,228,315,322,115 Neveh-Benjamin, M., 1391 141, 1J2 Newel, R., 270 Newell, A., 3 Newstead, S., 4Z Newtson, D., 6.8. Nicholls, W., ll, 289,290, 10.3 Nigam, A., 121 ll, 78, 852 951 96!

134

Copy righted m a rial

•

388

Author Index Nirruno-Smith, I., 84,115, W Nisbett, R., 111, lH Noelle-Neumann, E., 241-243,252,

253, 293, 30i Norman, D., 308, ill Nottenburg, G., 62 ovick L , ill ..

...

O'Brien, D., 176, 204 o·eonnell, c., 265,268,269,271 Oksenberg, L., 5,� 16,85,92, 94,

)Oil 314 Oldendick, R., 70,246 Olofsson, A., ill O'MaUey, P., �)69. 270 O'Muircheartaigh, C., 150,1561 2661

299,300,301,322,323 , O Reilly, j., 271, Z81, 300,302, 308 o·ReiUy, R., Z6 Ornstein, R., ill Osherson, D., 226 Ostrom, T., 214 Ottati, V., t "79t 201� .222 '

Pollack., L., 216 Poulton, E., 1171 244, ill Powell, M., 167,168, 179,ill Poynter, W., 11.8 Pratt, J., 281 Pratt, W., 233. 234,235,263,264,

265. 271,281,291) 297,298, 299,300,30.1 Pratto, F., 162. ' s., iU 22, 39, . 40, 43,60, 170, 174, 175, 197, 200, 201, 202,203,210,215,216, 2731 274,275, 295,296, 307,315, 328,330,331,332,333,337 Priester, J., 187. Prohaska, V., 116,121, 1.32 Psotka. J., 2.45 Puskar, c., 66 '

'

..

0

.

'

'

0

'

'

Quadrel, M., 1.61 Quigley, B., 271 Quillian, M., 77, 325

-

Padian, .· ··., 276 Panter, A., 67, 114 Parducci, A., �12, 214,220,;2232

239 Park, B., 7,. 1731 rn Patrick, D., 269,271 Payne, J., 8 Pelham, B., 112 Pepper, S., 41 Pergamit, M., 270,280,296 Perry, C., 265, 268, 269, 271 Person, N., 29, ll Petroni, R., 122 Petty, R., 7, 167, 187,�13, 7..211

2521307 Phillips, L., 16.1 Phipps, P., 290 Pierce, C., 271 Pierce, J., 269 Pijper, M., 290 PiUemer, D., 67, 10,89, 114 Pleck, J., 256,276, 290,300, 30& Pless, -., 2Zl Polivka, A., 324,32S

Raden, D., 1Z2 Radford A., �6, ll Raghubir, P., . SO, �55,156, �59,

337 R.aiffa, H , 281 Ra111irez, C., 66 ..

F., 4 Rasinski, K., Z 141 20,22, �72, �75, -r ·

t

t76t. 1111179,1aol 1st, m 186, 187. 188,189. 190,191, 1991lOt,. 202,�031205, �06, 207,209,228,233,234,235, . 251,263,264,265, 271,273, 276,281,282,283,284,291, 297,298,299,300,301,325, ill Raymond, P., 1.62 Reder, L., 11! llZ Reeves, B., 302 Reichenbach, H., " Reiser, B., 78, 79,8.1 Rhinehan, E., 67,89, 114. Rholes, W., 207, 337 . . . A., 62 Riggle, E., 79, 201,222 '

'

'

'

.

Copyrighted rna ri

I

Author lades Rips, L., 10,31,46,69,79,81,83, 88, 89, 90, 91, 114, 116,137, 147,244, 337 Ritteuauer-Schatka, H., 208 Roberts, R., 41 Robinson, J., 70, 76, 89,90, 2.1 Rocher, S., 1J

N., 212 Rogers, S., 276,290,300, 308 Rokeach, M., 174 Rosch, E., 46,68, ill Rosen, N., 213, 216 R R., 290 Rosen, S., 287 Ross, M., 13, W Rothgeb, J., ill Rowe, B., l,t, 301 Royston, P., 19, 323 Rubin, D., 861 87, 1322 13.3 Sadock, j., 106,108, 234

Safstrom, M., 303 Salter, W., 12,43,147, 148,149, 1S1,326,331,336t337,340 Sanbonmatsu, D., 167,168 , 172, t73, 111,178, t79,247, ru Sanchez, M., 93,336, llZ Sander, J., 19, 298,319 Sanders, L., 1ZS Sanford, A., 47,!2 Saris, W., 290 Schab, F., ll8

Schacter, D., Tl ...,..... G., 1.3 Scllaefer, E., ll Schaeffer, N., 47, 79, 86,235,236, 237,2S8, 284� 286,319, 340 _,.h ...,., R., Z1 Schank, R., 32, 69, Ll Schechter, S., 327,. 331, 332 Schlaifer, R., 281 Schmid, L, 265,268,269� 271 Schober, M., �53,551 56,51,319, 340 Schober, S., SS,270, 2801 2,96 Schooler, j., 71, 87, 118 Schuman, H., 1, � 43, 174, 175, I 78, 97, 200, 201, 202, 203,

389

210, 211, 215, 216, 217, 286, 315,337 Schwarz, ., 18, 22, 53, 54, 150, lSS, 156, 159, 160, 170, 176, 1n, tso, 200, 201, 202, 203, 204, 20S,207, 208, 209,211, 212� 2 I 3, 216, 221, 222�, i2,7.3, 224,225,226,232,241-243,24 7, 148, 252, 253, 262, 288, 291, 305, 310, 311, 3 14 , 3.37 Scoon-Rogers, L., zzz Scott, B., 326, 331 Searle, J., 29,246 Sean, D., 240, 241 Segal, G., ll ·

Seta, J., 1so, ,�.ot 222 Sewell, D., 6j Shafir, E., 2�2-6 Shahar, 1., llB Shapuo, G., 14,256,281 Sheatsley, P., � 210,337 Sheingold, K., 62 Shepherd, M., 216 Sherif, M., 3�37 Shem1an, S., 168,203! ,271.1,222, 337 SheveU, S., 10, 66, 79, 81, 83, 88, 89, 90, 91, 114,116,137, 147, 244,3JZ Shiffrin, R., 82 Shi mizu, 1., m Sboben, E., 46, 62 Shuan, M., 114, � Shyrock, H., 233 Sicoly, F., 13 Siegel, A., 2.i Siegel, J., ill Siegler, R., 337 Sigall, H., 269 Sikkel, D., 86. Silver, B., 273! 274, 280, 28Z Simon, H., 73,. 250! 326,334 Simons, A., 208 Sinclair, R., 146, 1491 150,152, 156,

278 Singer, E., 259,26� 263, 279, 288 Singer, M., 33, 4.1 Sirkin, M., 281,282 Skinner, S., 261,307

Copy righted m a rial

39 0

Author Index Skowronski, J., 11,66, 91,92,112,

114,,115, 116! 12 t 133. 244 Slovic, P., 95, 161,162, W Smith, A., � 12,11, 70, 85, 86, 97, 126! 148! H.2 Smith, E., 46, 69, 701 148, 2M Smith, K., 82, 85,93 , ;iJ6,ill Smith, T., 150t 154,169,175,199, 20.3,213,215,216,228,233, 234,247,263,264,268,269, 273,276,277,�80,�87,290, 294, �961 297,298,299,300, 301,305, 306 Sonenstein, F., 256,276,290,300, �108 Sperber, D ., 9,SA Sproull, L., 290,2991 301,. 102 Srull, T., 20Z Stalnaker, R., � Stapel, D., 180,. �13,221 Stcfanowska, M., 273 Stember, C., 286 Sternberg, S., 325 Stevens, S., 21S Stinson, L., :24! 264, 268,269,�72, 273,275,296,ill Stockwell, E., 2,33 Stokes, L., 320 Stokhof, M., 22 Stolley, K., 276 Suack F., 7, 17, 8, 20,53 , 150, 172,176,178,180,200,201, 202,203, 204,205,207,208, 209,211,212,220,223,226, 247, 248,310,311,314,319, ill Straf, M., 313, 335 Stroh, P., 173, 174 Stulginskas, J.,2Z3 Subramani, M., 302. Suchma� L., 54,57. 311,319, ll8 SudnJan, S., 21 22, 42,472 60, �5, 86, � 7! 88,89,92,94, �29- I 30, -32, �33, 45,146,147,151, 224, 7,25,251,259,260,268, 269, 280,307, 314,315,.llS Surnana, T., l12 '

'

.

..

' s., 32 Swa� G., 94,95,33St lli Sweeney, D., 21Z

-

.

·

-

'

Tan, E., 26i Tanfer, K., 256

Tangney, j.,285,306 T M., lZ Tanu.r, j.,313,ill Tenney, Y., 61 Tesscr, A., 171,184, �85,287 E., 11.8 Thomson, D., 80 Thompson, C., �s, �7,91, �2, 112t 114,115,116. 121,133,244 Thompson, D., 269, 271 Thornberry, 0., 293 Thornton, A., 335 Thurstone, L., � Tifft, L., 276 Tortora, R., llO Tourangeau, R., 7, 9, 2, 4. 7, 20, ·

.

22,43,51,66,97,146,147,148, 149, 150,.151, 152,'54, l72l 176!. t77, 179, t so•. 1st, m 186,187,188,189,190,191, 199, 2011 202,2031 205, �06, 207,217, 233,234,235,247, 250, 263,264, 265,269, 271, 273,2761 2771 281,290,�91, 296,297,298,299,300,301, 305,, 3131 314,319,325, 326, 331,335,336,337.�38,340 Trabasso, T., 3l Traugott, M., 274 Tripi� T., 307 Trope, Y., 203 Tsal, Y., ll8 Tuchfarber, A., 170,246 Tuchin, M., � Tulving, E., ?1-73,. 75,78, 80 Tupek, A., 290 Turner, C., 203, 256, 270,276,280, 290,296, 300,302,308, 313 Turner, T., 62 Tversky, A., �0, 421 119,. 125,1371 '

I

I

Copyrighted rna ri

I

Author Index 139,143, 144, 145,151, 160, 161,162, 163,ill Udry,J.,276 Underwood,B., 138, 139 Underwood,G., 118 Upchurch,D., 2_76 Upshaw, H., 214 Usher,j., 67, Z1 Vaughn,C., 269, 270 Verreault,R., 273 Von Neumann,j.,28.1 von Thurn, D., 262,288 Wadsworth,J.,lli Wagenaar,W.,11,75, 84,87,92, ~

Wagner,S., 217 Waksberg, j., 11 , 86, 88, 89, 92, 97, 126, 127, 128, 129-130, 133, 14b, ISS, 228,315, 322,ill Walker,J.,302 Walker,W., 121 WaUin, P., 262 Wallsten,T., 47, W Wanke,M., S3s 202,203 Warner,S., 272 Warriner,G., 2169 Waterton,j.,269, 301 Waterworth,j., 117 Watkins,M.,142,145, 1i1 Way,L., 276 Wedell, D., 212,220,223 Weeks,M.,290 Wegener,D., 213, 211 Wei.stnan,C., 226 Wellings,K., 276 Wells, G., 2121 1Uebllak,E.,264,268,269,2721l21

West, K., 281 Wetzel,A., 8Z Wetzler,S., 86, 8Z Whisman,M.,276 White,A., 93, 336, ill White,R., 1.H:

3 91

White,S., 67� 70, 89, 114, 116 Whitehouse,K., 1A3 Whitten, W.,83, 84, 2J Whittlesea,B., H.3 Wible,C., 8.1 Wickelgren,W., 86 Wiggins,S., W Wilkins,A., 112 Williams,C., 167,168, 171, 27�6 Williams,L., 271 Williams,M., � Willis, G., 19, 23, 14,6 .. 149 , 152, 258,281,282, 283,284,323, 327, 328, 330,331, 332,339 Wilson, D., 9, SA Wilson,T., 112,144, 16?,171, 177, 179, 185, 194, 197, 207! 208, 334 Wish, E.,270 Wittlinger,R., 83,� Woodberry,R., 21�S Woodrow,H., ll8 Wright,D., 150, ll6 Wyer,M., 285,287 Wyer,R., 80,179, 201, 207,213, 216,,,,, Wyner,G., 270

Xu, Y.,270 Yarrunarino,F., 261,307 Yates,S., 171, 1M Young, M., 274 Young,N.,123, 125, 1.26 Young-DeMarco,L., 335 Yzerbyt,V., U Zacks,R., 139, 140, at Zajonc, R., 241 Z,akay,D., 117,11.8 Zaller,J.,170, 172, 174, 175, 179, 185! 186, 187, 188! 189, 1.20 Zaragoza,M.,42 Zarrow,M., 12,11,78, 85,95,96, LM

Zhao,S., 2.95 Zitmnerman,J.,138, 112 Zwick,R., W

Copyrighted rna ri

I

abortion reporting� 265,268,271212 ACASI (audio computer-assisted self­ administered interviewing), 2561 276, 290,300,lCM 0

accessibility of attitudes,167-169, 179-180 of attitudinal considerations, 206202 of episodic infor1nation, 52, 155,

ill of nonepisodic information,155156, 159 accuracy of survey responses (see also measurement error),2 aggregate compari ons, 266· 0268 attitude questions, 2, 165-166 factual questions,94, iS frequency questions, 141, 149,

160 individual comparisons,266 26�8 measures of accuracy,�� .266-269 self vs. proxy responses,62. sensitive questions,264-265, 269-

.279 acquiescence, S additive decomposition,146, HZ adjunct questions,37-38 adverbial quantifiers, � age heaping,133-134 ambiguity setnantic, 2A syntactic,23-24, 35, �

I

•

I

I

anchoring-and-adjustn1ent,124,12SJ261 �44, �51,ill anchon,214, 232,239, 24S-246 argun1ent questions, 37-38 Anderson's information integration theory, 11, 13s �81, 2l1l ...... .

·

as�.

·

intelligence (AI), �62 110 lizecslaspcct markers, 101

asshnilation effects, 207, 247-248 attitudes crystallization,13,215 importance,.ill instability, 169 170 intensity,ill

strength, 117

attitude judgments-context effects,

197-219 attitude questions,answers to (see also belief-sampling model),165198 automatic processes, 168-169 basis for answers, 172-173 considerations, .&..a!!� ...-­

construal model, 167:-.168 deterntinants of response strategy,

177-178 effects of thought, 170 171 file drawer model,167, 112 traditional view, 166 167 0

attitude-behavior correlations, 1Z1 autobiographical events, 62· 63 autobiographical memory, 65,67-99. contents,68-71

392 Copy righted m a rial

Subject Index Conway's model, 75-77, Z2 extended events, 70=71 generic events, 69-70 Kolodner's model (CYRUS), 73-

75� 8.1 lifetime periods, ?G--71 retrieval from, 81-99 structure, 71-76 Tulving's model, 71-73 audio computer-assisted self­ administered interview (ACASI),

256, 276, 290,300,304 audio self-administered questionnaires (ASAQ), �911 304 availability heuristic, l37, 143-144, ill averaging model, see Anderson's in. . forrnauon mtegranon th •· eory .

backtracking, 15, 1i backward telescoping, se� telescoping balanced questions, 39 basic object categories, 6i behavior coding, 331-332 ling model, 178-194, 197belief·

198,22S-226,l23 judgment, 180 181 retrieval, 179 ·180. tests of the model, 185-194 bias models, 320 321 noaresponse, 3 ,... 2111.M. .O response propensity, ill biased rounding rules, 238-239 pipeline, 265, 268, 271, 2nm bounded interviews, 89-91, ll�t,ZB bounding, 89--91, 115-117 '

calendar prototypes , 234 Cannell, Miller, and Oksenberg model, S-7, 1�17, 314-315 card sorting, 323-324 ve Aspects CASM movement ( of Survey Methodology), 20-23, 0

0

·

·

313-317, 3 1 '. to progress, 337-340

'

·

·

and conceptions of survey measure­ ment error, 314-315,318-319 impact on psychology, 335-337 other effects on survey practice,

335

.

'

39 3

questionnaire development and testing, ;323 334 categorical response options ($ee also satisficing), 249-250 censoring, 256, 273, 276. 2,79 census participation, 262-263,279 channel of presentation (aural vs. vi· sual), 252, 292-293, 298, 300JOj childhood amnesia, Zl closed-ended questions, 382. 230-

2.11 coding schemes for cognitive inter· views, 327-3218 cognitive burden, ;J02-303, 305 cognitive interviewing, 326 328 coding schemes, 327-328 concurrent thlnk-alouds, 327 evaluation of, 331-333 paraphrasing, 327 probes, 32Z protocol analysis, 334 reliability of results, ill retrospective think-alouds, 327 cognitive reference points, �45, 2148 cognitive sophistication, 2752 cognitive toolbox, 8, ll components of the response process,

s, 7-16, 315-318 comprehension, 7-92 ,23-61 ambiguity, 273 24 complex syntax, H flexible interviewing, 57-59 inunediate understanding, 30-s34 interpretation, 30, 3,·i 34 logical for·rn. 31-32. pragmatics, ll role of inference, 31=34 semantics, 25, 40 SO standardization, 56 51 syntax, �5, 34 40. unfamiliar terms, 24, 43 '

Copyrighted rna ri

I

3 94

Subject lnde� comprehension (cont.) vague concepts, 24, 45·47 vague quantifiers, 47-50 computer assistance, 289-293, 299-

302

unconditional, ill contrast effects, 201, 202-205, 212•

215 conversational cues, 246-248

design principles, 302-312.: Cou­ per's, 309; Jenkins and DiU­

�onversational implicature, i1 · conversational see Grice•s conversational maxir11s Conway's model of autobiographical memory, 7S-772 1!1

man's, �09-310; Norman•s, 308302 hun1anizing the interface, .301-302 impact on reporting, 299-301 impersonality, 306-,JOZ legitimacy, 307-308 mental models, 311-312 mode of responding, 298, 302.-305 virtual human presence, 301-302. computer-assisted personal interview­ ing (CAPI), 276!. 289-290, 300,

cooperative principle, s�e Grice•s con· · versationaJ �· correlation between fonn, meaning, and use, '-9-30 cued recall, 84--86 Current Employment S urvey, 290 Current Population Survey (CPS), 56, 641 651 1011 108, �55, 258, 263, 268, 272, 289, 321, 322, 324 CYRUS (Kolodner's autobiographical memory model), 73-15, 78-79

liM computer-assi ted self-interviewing

decennial census, 262, 279.

cognitive burden, �02-3031. �051

308-310

(CASI), 276, 289, 300, 3M computer-assisted telephone inter­ viewing (CATI), 290, 300, 3M confidentiality, 259, 261-263, 279 considerations, see attitude questions, an wer to consistency, see editing consistency effects, see assimilation effects constant wave response, 125-126 construal model of attitudes, 167W Consumer Expenditure Swvey (CE), 1, 35, 8 , 101, 106, ill context effects (see also assimilation effects and contrast effects), 17, 20, 171� 198-229 and causal judgments, 2.12.-21J belief-sampling modei,Z15-2.2;6 conditional, 122 correlational, 198 directional, 198 frequency of, 215-217 inclusion/exclusion model (Schwarz and Bless), 221-2.25 '

'

·

0

declarative sentence, 29, 15 decomposition, 95-96, 162 deliberate misreporring, �e censoring depth of processing, 220, 307 design principles, see computer assis· tance Detroit Area Study (DAS), 215 diary studies, 66 67, 70, � disclosure to third parties, �59, 27928.1 to other government agencies, 259,

.279-281 within the rcspondent•s household, 258, 279-Z82_ discrimination net, ZJ disregarding accessible information,

'08-209 mood, ltt-212 j

distinctiveness of events, � Drug Use Forcasting (DUF), 270 duration questions, 102-103, 105-

W editing (see also censoring and misre­ porting), l, 13-14, 255-288

I

Copyrighted rna ri

I

Subject Index consistency, 2.87 interviewer approval, 257,275,

39 S

extrapolation, lA1 extreme exemplars, 214

279, 286 nlisreporting, 264-265, 269-279 and nonrespoDJC, 261-264 •

273 °275

politeness to interviewers, 2786 287 processes responsible for, 2.79-285 and reports about sexual behavior,

275-278 underreporting, 16�273 effects of thought on answers to atti­ tude questions, 170-171 elapsed ti111e questions, 102-103! 105107 embarrassment, 279, 282, 284-286, E-MOPs (event memory ! on packets), 73-75,78, 83 encoding, 139,235,317 episodic enumeration, see count episodic memory (Tulving's model), •

•

71-73 esti1nation for frequency questions (see aho l-and-count), 21, 88,143-145,147, �48-150,337 additive decomposition, 146, 147 ., on general impression, �41, 146,147, 149 150 based on generic information, �46, 147, 148-149 direct estimation, 149-150 exact tally, 1461 W recall-and-extrapolate, 146, 147148, 1.ll rough approxianation, 146,1491,50 strategy selection, 1S2�ts9 event series, 101-102. event tirne, 63·64 event-spetcific knowledge, Z1 exact tally, ill expert panels, 331-332 extended events, 7Q-71 calibration of probability judgments, 1.61 ·

.

55-57

306

�

factual questions, 1-22 �-122 61· 63 false alarm rate, l67 fan effect, 2l faulty presuppositions, 25,. 4 (........ file drawer model (Wilson and Hodges), 1671 172, 194 filled-duration illusion, 118-120 filter questions, � first-hand events, 6S�7 Fischhoff's panial perspectives approach, 181 ru flashbulb memories, &1 flexible vs. standardized interviewing, focus groups, 23� 327 focus, sentence, 2S forgetting, 82--91 length of reference period, 86-88 passage of time, 82-86 proximity to temporal boundaries,

88-91 forgetting curves, 86-88 forward telescoping, see telescoping frequency tes, see estima tion for frequency questions frequency of context eff� 215-217 frequency questions, see factual ques­ tions; estimation for frequency quesnons ·

•

Gallup Poll, 174,lli Galton method, 8Z general happiness-marital happiness,

20.YOS

general political values, J74 General Social Survey (GSS), � 13,

23,25,62,101,215,232,251 general-specific questions, 203 205 generic information, 69-70,71,1481A2

generic memories, 21,69-70, 7& ·79. Gigerenzer and Goldstein's take-the­ best heuristic, Hi gra1n1nar, ue syntax

Copyrighted rna ri

I

396

Subject Index Grice's conversational maxims, 5153, 202-203, 2 205, 209 cooperative principle, li maxirn of rnanner, .U maxim of quality, i1 maxim of quantity, 51,204, 20i maxirn of relation, 51,202 gross discrepancy rate, 266 grounding, ll I

interview tirne, 63-64,30.1 interviewer, h 5, 276, 286, 292,294195,297-298 interviewer approval, 257, 275, 279,

2186 interviewer debriefing, 331-332 inuusive questions, 255-256,258, 261 itetn nonresponse, 260-261, 263264, 2731 299

Hansen-Hurwitz-Bershad model,319-

320 high road-low road theories/two track theories, l&-.19 Hintz1nan and Curran theory of rec­ ognition judgments, 142-143

judgments in surveys (see also estima­ tion for frequency questions and attitude judgments), 7-8, 10 ,13 judgtnental contrast effects, I 2 1 judgmental heuristics, �37, 143-145, I�

111 ideological predispositions, m illicit drug use, 270, 294-296 inunediate question comprehension, 30c3l imperative sentence, 27-2.8 implica tures in survey� ll!e also con­ versational cues), 53-SS, 246· ·

anchoring-and-adjustment, 124, 125-126,144,151,247 availability heuristic, 137, 143144, ill representativen(SS heuristic, 143,

ll1 judgments of causality, 212-213

248

'

implicit memory, 77-78 impression-based judgments, 142�43, -49-1 0, ,73-174, 247 for attitude questions, 173-174 for frequency questions, 142-143, 149-150,247 inclusion/exclusion model for context effects (Schwarz and Bless), 221-

225 income reporting, 263-264, 268, 272. inductive inference, 226-2278 inference, see comprehension; reconstrucuon instability of attitudes, 169-170 interaction coding/behavior coding, •

l26 interactive voice response (IVR), 290, liM interference effect, 83 Internet surveys, 290 interpretation, see comprehension

interrogative sentence, 27-29. 342 38

Kolodner's model (CYRUS), 73-75,

83 K.rosnick and Alwin's model for re­ sponse order effects, 17,250-

154 strong satisficing, 253 weak satisficing, m lack-of-knowledge inference, 43, 157 landmark events, 67, 70-7l, 79-80, 89-91, 113-115 leading questions, 42 length of reference period, 86 88 leniency bias, 240-241,ill level of item generality, ��2091

22_6-2,27 assimilation effects, 201-203 conttast eff�,201,202-20S,212215 life event calendar, 91, 335-336 life satisfaction, judgments of, 211-

211

I Copy righted m a rial I

Subject Index lifeti&Jle periods,70-71, 75-761 7980 logical form, rll-32 long-term memory, Tl Iymg,26S,279,284-286 zr1apping of response, 13-14, 232235,239-249 mean absolute difference,266. squared error,266 measurement error (see also effects),4, 121-122,265-267, 276,340 342 random error,266 systematic error,2.66 medium of questionnaire (paper vs. electronic),292-293, 298-302 memory-based judgments,7,21 memory failure,see forgetting memory for ela titne, 110-111 memory indices,see £-MOPs Memory tion Padcets (MOPs),69,73-75 method of ad1ninistration (self vs. in­ terviewer; see also self­ a�uusttation),294-298, 300 method of contact,293-295 misreporring, 264--265,269-185 ·

397

viewing (CAPI), 276,. 289-290, 300,304 computer-assisted self-administered interviewing (CASI),276,289, 300,304 computer-assisted telephone inte_r­ viewing (CATI),290,300, 304 disk by mail (DBM),290, o/104 interactive voice response (IVR), 290, 304 Internet surveys, 290 paper-and-pencil penonal inter­ viewing (PAPI),290-291,294, 300, 3M prepared data entry (POE),290, 1M self-administered questionnaire (SAQ),250, 2651 27o-271, 275, 279,291,295,297-298,300,

.104 touchtone data entry {T'DE),290, 304 voice recognition entry (VRE),290� 3!M mode of responding,298, 302-,305 Monetary Conttol Bill,175, 202-

203 mood effects,211-212

avoiding e1nbarrassment,265,284-

285 confidentiality,27�181 lying, 265, 279,284-286 privacy,265,279 2,80 question threat,264 missing data,see item nonresponse miss rate,267 mixed views,attitude question&,186-

188 mode of data collection,20, 276278,289 312 audio computer-assisted self­ adrninistered interviewing (ACASJ),256, 276,290,300, 304 audio self-administered question­ naire (ASAQ),291,304 computer-assisted personal inter-

National Crime Survey (NCS),1,11., 100� 101s 104, ill National Education Longitudinal Study of 1988,2nd Follow-up, 261-263 National Election Studies (NES), 186, 2371 2731 280 National Health Interview Survey (I-US), l21 38, 62, 63, lOl, 103, 136,165, 2311 3111 3131 324 National Household Survey Of Drug Abuse (NHSDA),256,261 National Longitudinal Survey of Youth (NLS-Y),282 National Medical Expettditure Survey (NMES),289 National Opinion Research Center (NORC), 174

Copy righted m a rial

398

Subject Index y Growth National Survey of ( SFG), 26St 282 need for social approval, 258 ·

net discrepancy rate, 266-267 nonatritudes, 1.6.2 norrn of evenhandedness, 211 nortn of politenesst 286.287 ntmleric labels, 230, 241-244 nurnerical reference points, 24\tS numerosity heuristic, 112. on-line judgments, Z open-ended questions, 175-176, 186187, 2131-232 optimizing, 25 1 overlapping consideration$, 188:190 overreporting of desirable behaviors,

273-225 voting, 273-274_, 278 church attendance, 274-275 pace of administration, 310-311, 333-

334 Panel Study of Income Dynamics (PSID), 122 paper-and-pencil personal interview­ ing (PAPI), 290-291, 294, 300, 31M partial prespectives, 173 175 passage of time, 82.-86 personal narratives, Zl Petty and Cacioppo's theory of per­ suasion, 2:.52-253 phonological strucnare of questions,

26

•

I

politeness to interviewen, 286-r287 interviewer class, 2-86, 2.87 interviewer race, '�86 �187 interviewer sex, 286 287 positional cues, 247-248. positivity bias (see also leniency), 240245, 148 pragmatics, 25, 51-56 of interviews, 54-56 prepared data entry (PDE), 290, 3M presupposition, 25, 41� 44

pretesting methods, 323 328 ...... sorting, 32.3- 32.4. cognitive interviews, 23, 326- 318 ·

confidence rating&. ill expert panels, J2Z focus group discussions, 23! ill latency, llZ

324 -t3lS prunacy e ; see response order effccts priming effects, 176 177 •

•

privacy, 258, 259, 263, 275-276,

21!)_ probability judgments, 160-165, external calibration, 1.6..1 interual , 16l-162: conjunction effect, 161; disjunctive events, 162; un 162 protocol analysis (ue 11/so cogtlitive interviewing), 3M prototypes, 46, 228 calendar, 234 numerical, 245 prototypical question, 30 proxi1nity to temporal boundari� 88.... � ..

·

·

·

2..1

proxy reports, 12, 65 61 psychological continuun1, 4, S. psychomettic theories of the response process, H quality profiles, 319 question comprehension, 30 31 irnmediate understanding, 30=31 interpretario� 30-31 question context effects, see context effects question order effects, see context ef­

fects question wording, 2J, 44 adverbial quantifiers, 24, 47-50 focus, se11teoce, 25, 32, 35-38 seanantic , 23-24 , 23-24, 34-38, syntactic 4Q unfamiliar terms, 24, 42-43 ·

·

·

·

I • I

Copyrighted rna ri I

I

Subject Iadex vagueness, 24, 45-SQ question wording e ffects, 174 175 Korean war items, ____ .....,..

rando response technique, 272.2478 range-frequency model, 214--215, 23�241, 244, ,MS frequency principle, 240 242 range principle, 214-215, m rating 241-246, ill rational deliberation and misreport­ ·

.

ins, 279

reaction time, 167-1692 325-326 attitude questions, 167-169, 325 recaU, 81-99, 102 recall order, 23-94 recall-and-count (episodic enumeration), 146, 151, 153-156, 158160, ru recall-and-extrapolate (rate estiJJJa­ tion), 147-148, m recency effects, se� response order ef­ fects reconstruction, 12, 81-82 reference date, 5 reference period, 11, 64-65, 86 reference points, 245, 248 relative temporal order, 112�-113 reliability of answers to attitude questions, see response stability reprcscrttation-about the sentence, 310

M

represesttation-of the sentence, 31y

representativeness heuristic, 143, W response aids, 335-338 response contraction, 1 7- 18., 2 245,248 response effects, 2-31 8, 12 ts, 230::232 IUponsc order effects, 250-254, 304-

lOS and channel of presentation, 3 305,251-252 cognitive sophistication, 2.52.

·

399

individual differences, 252 Krosnick and Alwin's satisficing model� 251-252 primacy effects, 251-252 recency effects, 251-252 models of, 2-72 16response 19, 41, 315-319 Cannell, Miller, and Okse�tberg model, S-7 high road-low roadltwOo-track models, 16 ·19 Krosnick and Alwin's satisficing model, 17, 251-252 psychometric models, bZ Strack and Martin •s model for atti­ tude questions, 17-19. response stability for attitude ques­ tions, 181-184 retrieval, 7-8, 9-10, 77-81 relation between retrieval and judg­ ment for factual questions, 9U retrieval cues, 78-811 96-97 relative effectiveness of different cues, 78-79 retrieval·based assimilation effects, 206-207 retrospective bias, 125-126 retrospective probes, ill rounding, 22, 162-163, 232-23� encoding, 2ll feeling thermometer, 237-138 indeterminate quantities, 2r1S-238 magnitude of quantity being estimated, 234-235 roWiding rules, 238-239 ·

0

sampling errors, 2. satisficing, 17, 2SG-251, lOS scale anchors, 245-2.46 labels, 241-245, 248 scale range effects, 249 scale values, lll-2113 Schwarz and Bless's inclusion/exclu­ sion model, 2.11-225 scripts, 69-70

Copyrighted rna ri

I

400

Subject Index seam effect, 122-126 constant wave response, 125-ll6 forgetting, 1'-477126 retrospective bias, 12S-126 second-hand events, 6.S 67 tion of events, 118=112 self-adxninistrarion, 265, 270-271, 275-276.!.79,282,295,297298 300 310 self-presentation, 278s$291 semantic memory, 72-73 semantic problems, 23-ZS '

'

.

'

sensitive questions (threatening questions), 255-288,291 sentence parsing, 36, lZ serial position effe� 228-229 sexual partners, �32-234, 268-269, 275-278, 296-1797 show cards, 250-293 simple response variance, 322-323 sincerity conditions, 246 social desirability, S, 257, 294, 296, 3fU Soc:ratic effect, 1�13 space of uncertainty, 27-30, 32, 34, 4�50,246 spread of activitation, T1 standard of comparison, 2 0, 212 standardized interviewing, SS-57 statistical models of error, 319-321 Strack and Martin's two-track theory, 17-19 subjective expected utility theory (SEU), 181-284 risks and losses, 2.81-282. subtraction-based contrast effects, 205 Sudman-Bradburn theory of time compression, 129 13l summarized events, ZB Survey of Consumer Finances, 234235! 289

Survey of Income and Program Par­ ticipation (SIPP), 122 Survey on Census Panicipation (SCP), 263 syntax, �3-24, 25, 34, 38

syntactic

·

biguity, 24-25, 34-38,

~ syntactic probleans in question wording, 34-50 telescoping, 11,8�89, 120-121,126132, 335-331 '

backward, �8-891 12Jh121 forward, 11, 88-89, 12Q..-12.1 inter11al, 127-118 Neter and Waksberg's study, 126129 temporal compression, 128-132 variance theories, 132.-133 temporal boundaries, 89-91 temporal compression, see telescoping temporal frequency questions, 102103, 107-108 temporal landr11arks, 89-91, I13-llS calendar based events, 89-91, 1.1A landmark events, 89-91, 1.H temporal periods/sequences, 11S-117 temporal questions constraint-satisfaction procedure, 110 112 impressions based on reuieval at­ tempt, 109 and recall of exact temporal infor­ mation, 1.!l2 and recall of relative order infor... mation, 1091 113-117: extended event, 109, 115-117; temporal landmark, 09, � 13-115; tempo­ ral period. 115-117 types of temporal questions, 101108: questions about duration, 105-107, 117-120; questions about elapsed tinte, 105-107, .20-121; questions about teua­ poral rates, 107-108; time-of­ occurrence questions, 104-105, .

.

112-117 temporal rates, 107-108 third parties, see confidentiality time on task, 941 9�i time-of-occurrence questions, 102lOJ, 104·-lO.S

Copyrighted rna ri

I

Subject Index topic saliency, 261 Touchtone Data Entry (TOE), 290,

304 trace position, 36 trace location process, 36-37 traditional view of attirude questions,

165-167 Tulving's model of episodic memory,

71-73 Tversky and Kahneman's judgmental heuristics, see judgmental heuristtcs typicality effects, 49 ..

unit nonresponse, 261, 264, 273, 301 unpacking effect, 119 U.S.-Conununist reporters, 210-211 vague concepts, 45-47 vague quantifiers, 47-50 vagueness, 24, 44 50 valuation process, 182 variability of understanding, 4S variance theories of telescoping, 132-

133 verbal labels for response options,

230 verification questions, 103-104, 107-

unbounded interviews, 126-128 undercoverage, 256, 281 underreporting of undesirable behaviors, 266-278 abortion, 264, 269, 271-272 consumption of alcohol, 269 use, 269-270, 295-296 illicit racist attitudes, 269 smoking, 269, 271, 278 unfamiliar terms, 24, 43

I I

I

•

I I

t

108 Vierordt's law, 118-119 vignettes, 282-284 Voice Recognition Entry (VRE), 290,

304 wh-questions, 35-38 working memory, 36-38, 40, 77

ZUMA, 314

40 1

ISBN 0-521-57629 - 6 -

-

-

-

-